Hindawi Publishing CorporationJournal of Electrical and Computer EngineeringVolume 2013 Article ID 837275 12 pageshttpdxdoiorg1011552013837275

Research ArticleOnline Detection of Abnormal Events in Video Streams

Tian Wang1 Jie Chen2 and Hichem Snoussi1

1 Institut Charles Delaunay LM2S-UMR STMR 6279 CNRS University of Technology of Troyes 10004 Troyes France2Observatoire de la Cote drsquoAzur UMR 7293 CNRS University of Nice Sophia-Antipolis 06108 Nice France

Correspondence should be addressed to Tian Wang wangtian8704gmailcom

Received 19 September 2013 Accepted 12 November 2013

Academic Editor Yi Zhou

Copyright copy 2013 Tian Wang et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

We propose an algorithm to handle the problem of detecting abnormal events which is a challenging but important subject invideo surveillance The algorithm consists of an image descriptor and online nonlinear classification method We introduce thecovariance matrix of the optical flow and image intensity as a descriptor encoding moving information The nonlinear onlinesupport vector machine (SVM) firstly learns a limited set of the training frames to provide a basic reference model then updates themodel and detects abnormal events in the current frame We finally apply the method to detect abnormal events on a benchmarkvideo surveillance dataset to demonstrate the effectiveness of the proposed technique

1 Introduction

Visual surveillance is one of the major research areas incomputer vision In a crowd image analysis problem thescientific challenge includes abnormal events detection Forinstance Figure 1(a) illustrates a normal scene where thepeople are walking In Figure 1(b) all the people are suddenlyrunning in different directions This dataset imitates panic-driven scenes

Trajectory analysis of objects was described in [1ndash3] Themoving object was labeled by a blob in consecutive framesand then a trajectory was produced The deviation from thelearnt trajectories was defined as abnormal events Trackingbased approaches are suitable for the sparse scenes with a fewobjects The target might be lost due to occlusion

In [4 5] abnormal detection approaches which used fea-tures encoding motion texture and size of the objects wereintroduced Local image regions in a video were analyzed byemploying background subtraction method then a dynamicBayesian network (DBN) was constructed to model normaland abnormal behavior and finally a likelihood ratio testwas applied to detect abnormal behaviors In [6] a space-time Markov random field (MRF) model which detectedabnormal activities in a video was proposed mixture ofprobabilistic principal component analyzers (MPPCA) wasadopted to model local optical flow The prediction is based

on probabilistic assumption techniques where an accuratemodel exists but there are various situations where a robustand tractablemodel cannot be obtained model-freemethodsare needed to be studied

Spatiotemporal motion features described by the contextof bag of video words were adopted to detect abnormalevents In [7] the authors presented an algorithm whichmonitored optical flow in a set of fixed spatial positionsand constructed a histogram of optical flow The likelihoodof the behavior in a new coming frame concerning theprobability distribution of the statistically learning behaviorwas computed If the likelihood fell below a preset thresholdthe behavior was considered as abnormal In [8] irregularbehavior of images or videos was detected by an inferenceprocess in a probabilistic graphical model In [9 10] thevideo pixels were densely sampled to form the feature Thesemethods are based on the partial information of images suchas small blocks in a frame without fully exploiting the globalinformation of the feature In [11ndash13] spatiotemporal featuresmodeled motion regions of the frame as background andanomaly was detected by subtracting the newly sample to thebackground template These works are similar to the changedetection method when the background is not stable

In this paper the proposed algorithm is composed of twoparts Firstly a covariance feature descriptor is constructedover the whole video frame and then a nonlinear one-class

2 Journal of Electrical and Computer Engineering

support vector machine algorithm is applied in an onlinefashion in order to detect abnormal events The featuresare extracted based on the optical flow which presents themovement information Experiments of real surveillancevideo dataset show that our online abnormal detection tech-niques can obtain satisfactory performance The rest of thepaper is organized as follows In Section 2 covariance matrixdescriptor of motion feature is introduced In Section 3 theonline one-class SVM classification method is presented InSection 4 two abnormal detection strategies based on onlinenonlinear one-class SVM are proposed In Section 5 wepresent results of real-world video scenes Finally Section 6concludes the paper

2 Covariance Descriptor of Frame Behavior

The optical flow is a feature which presents the directionand the amplitude of a movement It can provide importantinformation about the spatial arrangement of the objects andthe change rate of this arrangement [14] We adopt Horn-Schunck (HS) optical flow computation method in our workThe optical flow of the gray scale image is formulated as theminimizer of the following global energy functional

119864 = ∬[(119868119909119906 + 119868119910V + 119868119905)2

+ 1205722

(nabla1199062

+ nablaV2)] 119889119909119889119910 (1)

where 119868 is the intensity of the image 119868119909 119868119910 and 119868

119905are the

derivatives of the image intensity value along the 119909 119910 andtime 119905 dimension 119906 and V are the components of the opticalflow in the horizontal and vertical direction and 120572 representsthe weight of the regularization term

We introduce the covariance matrix encoding the opticalflow and intensity of each frame as the descriptor to representthemovementThe covariance feature descriptor is originallyproposed by Tuzel et al [15] for pattern matching in a targettracking problem The descriptor is defined as

119865 (119909 119910 119894) = 120601119894(119868 119909 119910) (2)

where 119868 is the color information of an image (which can begray RGB HSV HLS etc) 120601

119894is a mapping relating the image

with the 119894th feature from the image 119865 is the 119882 times 119867 times 119889

dimensional feature extracted from image 119868119882 and119867 are theimage width and image height and 119889 is the number of chosenfeatures For each frame the feature can be represented as119889 times 119889 covariance matrix

C =1

119899 minus 1

119899

sum

119896=1

(z119896minus 120583) (z

119896minus 120583)⊤

(3)

where 119899 is the number of the pixels sampled in the framez119896is the feature vector of pixel 119896 120583 is the mean of all the

selected points and C is the covariance matrix of the featurevector 119865 The covariance descriptor C of each frame dose nothave any information regarding the sample ordering and thenumber of points [15] Because the feature 119865 can be designedas different approaches to fuse features the covariancematrixdescriptor proposes a way to merge multiple parametersDifferent choices of feature vectors extraction are shown in

Table 1 Different choices of feature 119865 to construct the covariancedescriptor

Feature vector 1198651198651(6 times 6) [119910 119909 119906 V 119906

119909119906119910]

1198652(6 times 6) [119910 119909 119906 V V

119909V119910]

1198653(8 times 8) [119910 119909 119906 V 119906

119909119906119910V119909V119910]

1198654(12 times 12) [119910 119909 119906 V 119906

119909119906119910V119909V119910119906119909119909

119906119910119910

V119909119909

V119910119910]

1198655(17 times 17) [119910 119909 119906 V 119906

119909119906119910V119909V119910119906119909119909

119906119910119910

V119909119909

V119910119910

119868 119868119909119868119910119868119909119909

119868119910119910]

Table 1 where 119868 is the intensity of the gray image 119906 and Vare horizontal and vertical components of optical flow 119868

119909

119906119909 and V

119909are the first derivatives of the intensity horizontal

optical flow and vertical optical flow in the 119909 directionrespectively 119868

119910 119906119910 and V

119910are the first derivatives of the

corresponding feature in the 119910 direction and respectively119868119909119909 119906119909119909 and V

119909119909are the second derivatives in 119909 direction

119868119910119910 119906119910119910 and V

119910119910are the second derivatives in 119910 direction

The flowchart of covariance matrix descriptor computationis shown in Figure 2 The optical flow and correspondingpartial derivative characterize the interframe information orcan be regarded as the movement information The intensityof the frame and partial derivative of the intensity describethe intraframe information they encode the appearanceinformation of the frame

If proper parameters are given the traditionally usedkernel such as Gaussian polynomial and sigmoidal kernelhas similar performances [19] Gaussian kernel 120581(x

119894 x119895) =

exp(minusx119894minus x1198952

21205902

) is chosen for our spatial features Thecovariance matrix is an element in Lie group the Gaussiankernel on the Euclidean spaces is not suitable for the covari-ance descriptorsTheGaussian kernel in Lie Group is definedas [20 21]

120581 (X119894X119895) = exp(minus

10038171003817100381710038171003817log (Xminus1

119894X119895)10038171003817100381710038171003817

21205902) (X

119894X119895) isin 119866 times 119866

(4)

where X119894and X

119895are matrices in Lie Group 119866

3 Online One-Class SVM

The essence of an abnormal detection problem is that onlynormal scene samples are available The one-class SVMframework is well suitable to an abnormal detection problemSupport vector machine (SVM) is initially proposed byVapnik and Lerner [22 23] It is a method based on statisticallearning theory and has fine performance to classify data andrecognize patterns There are two frameworks of one-classSVM one is support vector data description (SVDD) whichis presented in [24 25] and the other is ]-support vectorclassifier (]-SVC) introduced in [26]The SVDD formulationis adopted in our work It computes a sphere shaped decisionboundary with minimal volume around a set of objects The

Journal of Electrical and Computer Engineering 3

(a) Normal lawn scene (b) Abnormal lawn scene

Figure 1 Examples of the normal and abnormal scenes (a) Normal lawn scene all the people are walking (b) Abnormal lawn scene all thepeople are running

Consecutive frameFramei

Framei+1

Optical flow field OPi

Features

F(x y j)

j = 1 2 nCframe119894

Figure 2 Covariance descriptor computation of a video frame

center of the sphere c and the radius 119877 are to be determinedvia the following optimization problem

min119877120585119888

1198772

+ 119862

119899

sum

119894=1

120585119894 (5)

subject to 1003817100381710038171003817Φ (x119894) minus c1003817100381710038171003817

2

le 1198772

+ 120585119894 120585119894ge 0 forall119894 (6)

where 119899 is the number of training samples and 120585119894is the

slack variable for penalizing the outliersThe hyperparameter119862 is the weight for restraining slack variables it tunesthe number of acceptable outliers The nonlinear functionΦ X rarr H maps a datum x

119894into the feature space

H it allows to solve a nonlinear classification problem bydesigning a linear classifier in the feature space H 120581 is thekernel function for computing dot products inH 120581(x x1015840) =⟨Φ(x) Φ(x1015840)⟩ By introducing Lagrange multipliers the dual

problem associated with (6) is written by the followingquadratic optimization problem

max120572

119899

sum

119894=1

120572119894120581 (x119894 x119894) minus

119899

sum

119894119895=1

120572119894120572119895120581 (x119894 x119895)

subject to 0 le 120572119894le 119862

119899

sum

119894=1

120572119894= 1

c =119899

sum

119894=1

120572119894Φ(x119894)

(7)

The decision function is

119891 (x) = sgn(1198772 minus119899

sum

119894119895=1

120572119894120572119895120581 (x119894 x119895)

+ 2

119899

sum

119894=1

120572119894120581 (x119894 x) minus 120581 (x x))

(8)

For the large training data the solution cannot beobtained easily and an online strategy to train the data is

4 Journal of Electrical and Computer Engineering

used in our work Let cD denotes a sparse model of the centerc119899= (1119899)sum

119899

119894=1Φ(x119894) by using a small subset of available

samples which is called dictionary

cD = sum

119894isinD

120572119894Φ(x119894) (9)

whereD sub 1 2 119899 and let119873D denote the cardinality ofthis subset xD The distance of a mapped datum Φ(x) withrespect to the center cD can be calculated by

1003817100381710038171003817Φ (x) minus cD1003817100381710038171003817 = sum

119894119895isinD

120572119894120572119895120581 (x119894 x119895)

minus 2sum

119894isinD

120572119894120581 (x119894 x) + 120581 (x x)

(10)

A modification of the original formulation of the one-class classification algorithm that consists of minimizing theapproximation error c

119899minus cD is [27 28]

120572 = arg min120572119894119894isinD

1003817100381710038171003817100381710038171003817100381710038171003817

1

119899

119899

sum

119894=1

Φ(x119894) minus sum

119894isinD

120572119894Φ(x119894)

1003817100381710038171003817100381710038171003817100381710038171003817

2

(11)

The final solution is given by

120572 = Kminus1120581 (12)

where K is the Grammatrix with (119894 119895)th entry 120581(x119894 x119895) and 120581

is the column vector with entries (1119899)sum119899119894=1

120581(x119896 x119894) 119896 isin D

In the online scheme at each time step there is a newsample Let 120572

119899denote the coefficients K

119899denote the Gram

matrix and 120581119899denote the vector at time step 119899 A criterion is

used to determine whether the new sample can be includedinto the dictionary A threshold 120583

0is preset for the datum

x119905at time step 119905 the coherence-based sparsification criterion

[29 30] is

120598119905= max119894isinD

1003816100381610038161003816120581 (x119894 x119905)1003816100381610038161003816 (13)

First Case (120598119905gt 1205830) In this case the new data Φ(x

119899+1) is not

included into the dictionaryThe GrammatrixK119899+1

= K119899 120581119899

changes online

120581119899+1

=1

119899 + 1(119899120581119899+ b)

120572119899+1

= Kminus1119899+1120581119899+1

=119899

119899 + 1120572119899+

1

119899 + 1Kminus1119899b

(14)

where b is the column vector with entries 120581(x119894 x119899+1

) 119894 isin D

Second Case (120598119905le 1205830) In this case the new data Φ(x

119899+1) is

included into the dictionaryD The Grammatrix K changes

K119899+1

= [

[

K119899

bb⊤ 120581 (x

119899+1 x119899+1

)]

]

(15)

By using Woodbury matrix identity

(119860 + 119880119862119881)minus1

= 119860minus1

minus 119860minus1

119880(119862minus1

+ 119881119860minus1

119880)minus1

119881119860minus1

(16)

Kminus1119899+1

can be calculated iteratively

Kminus1119899+1

= [

[

Kminus1119899

00⊤ 0

]

]

+1

120581 (x119899+1

x119899+1

) minus b⊤Kminus1119899b

times [

[

minusKminus1119899b

1]

]

[minusb⊤Kminus11198991]

(17)

The vector 120581119899+1

is updated from 120581119899

120581119899+1

=1

119899 + 1[119899120581119899+

120581119899+1

] (18)

with

120581119899+1

=

119899+1

sum

119894=1

120581 (x119899+1

x119894) (19)

Computing 120581119899+1

as (19) needs to save all the samples x119899+1119894=1

inmemory For conquering this issue it can compute as 120581

119899+1=

(119899 + 1)120581(x119899+1

x119899+1

) by considering an instant estimationTheupdate of 120572

119899+1from 120572

119899is

120572119899+1

=1

119899 + 1[119899120572119899+ Kminus1119899b

0]

minus1

(119899 + 1) (120581 (x119899+1

x119899+1

) minus b⊤Kminus1119899b)

times [

[

Kminus1119899b

1]

]

(119899b⊤120572119899+ b⊤Kminus1

119899b minus 120581119899+1

)

(20)

Based on (20) we have an online implementation of the one-class SVM learning phase

4 Abnormal Events Detection

In an abnormal event detection problem it is assumedthat a set of training frames 119868

1 119868

119899 (the positive class)

describing the normal behavior is obtained The generalarchitectures of abnormal detection are introduced below

The offline training strategy refers to the case where allthe training samples are learnt as one batch as shown inFigure 3(a) We propose two abnormal detection strategiesthe difference between these two strategies is the time whenthe dictionary is fixed These two strategies are shown inFigures 3(b) and 3(c) Strategy 1 is shown in Figure 3(b) Thetraining data are learnt one by oneWhen the training periodis finished the dictionary and the classifier are fixed Eachtest datum is classified based on the dictionary Figure 3(c)illustrates Strategy 2 The training procedure is the same asStrategy 1 But in the testing period the dictionary is updatedif the datum x

119894satisfies the dictionary update condition The

details of these two strategies are explained in the following

Journal of Electrical and Computer Engineering 5

Train offline

Test onlinem n minus m

(a) Strategy offline

Test online

Train online

m n minus m

Train

Dictionary fixed

(b) Strategy 1

Train

Dictionary fixed

Train and test online

Test onlinem n minus m

(c) Strategy 2

Figure 3 Offline and two online abnormal event detection strategies based on one-class SVM (a) Strategy offlineThe training data are learntas one batch offline (b) Strategy 1 The dictionary is fixed when all the training data are learnt (c) Strategy 2 The dictionary continues beingupdated through the testing period

Features selectionon original image

Learning step (online)Covariance

Covariance

optical flow

Classification

SVM train

(online training)

people walk

Detection step (online)One-class SVM

R

Origin

Abnormal eventdetection

people run

optical flow

x11 x12 middot middot middot

middot middot middot

middot middot middot

x1n

x21 x22 x2n

xn1 xn2 xnn

x11 x12 middot middot middot

middot middot middot

middot middot middot

x1n

x21 x22 x2n

xn1 xn2 xnn

⋱

⋱

Figure 4 Major processing states of the proposed abnormal frame event detection method The covariance of the frame is computed

6 Journal of Electrical and Computer Engineering

41 Strategy 1 In Strategy 1 the dictionary is updated merelythrough the training period

Step 1 The first step is calculating the covariance matrixdescriptor of training frames based on the image intensity andthe optical flow This step can be generalized as

(1198681OP1) (1198682OP2) (119868

119899OP119899) 997888rarr C

1C2 C

119899

(21)

where (I1OP1) (I2OP2) (119868

119899OP119899) are the image

intensity and the corresponding optical flow of the 1st to 119899thframe C

1C2 C

119899 are the covariance matrix descriptors

Step 2 The second step consists of applying one-class SVMon the small subset of extracted descriptor of the trainingnormal frames to obtain the support vectors Consider asubset C

119894119898

119894=1 1 le 119898 ≪ 119899 of data selected from the full

training sample set C119894119899

119894=1 without loss of generality and

assume that the first 119898 examples are chosen This set of 119898examples is called dictionary CD

C1C2 C

119898 1 le 119898 ≪ 119899

SVM997888rarr support vector 119878119901

1 1198781199012 119878119901

119900

(22)

where the set C1C2 C

119898 is the first119898 covariancematrix

descriptors of the training frames it is the original dictionaryCD In one-class SVM the majority of the training samplesdo not contribute to the definition of the decision functionThe entries of a monitory subset of the training samples1198781199011 1198781199012 119878119901

119900 119900 le 119898 are support vectors contributing

to the definition of the decision function

Step 3 After learning the dictionary CD which includesthe first 119898 1 le 119898 ≪ 119899 samples the training samplesC119898+1

C119898+2

C119899 are learned online via the technique

described in Section 3 This step can be generalized as

CDC119896 119898 lt 119896 le 119899SVM997888rarr support vector 119878119901

1 1198781199012 119878119901

119901

119900 le 119901 le 119899

CD = CD cup C119896 if 120598

119905ge 1205830

(23)

where CD is the dictionary obtained through Step 2 C119896is a

new sample in the remaining training dataset According tothe criterion introduced in Section 3 if the new sample C

119896

satisfies the dictionary updated condition it will be includedinto the dictionary CD

Step 4 Based on the dictionary and the classifier obtainedfrom the training frames the incoming frame sample C

119899+119897is

Table 2 AUC of abnormal event detection method of differentfeatures 119865 via original SVDD which learns training samples offlineStrategy 1 online one-class SVM and Strategy 2 online one-classSVMThe biggest value of each method is shown in bold

Features Area under ROCLawn Indoor Plaza

Training samples are learned offline1198651(6 times 6119889119906) 09426 08351 09323

1198652(6 times 6119889V) 09400 08358 09321

1198653(8 times 8) 09440 08375 09359

1198654(12 times 12) 09591 08440 09580

1198655(17 times 17) 09567 08649 09649

Strategy 11198651(6 times 6119889119906) 09399 08328 09343

1198652(6 times 6119889V) 09390 08355 09366

1198653(8 times 8) 09418 08377 09411

1198654(12 times 12) 09581 08457 09573

1198655(17 times 17) 09551 08628 09632

Strategy 21198651(6 times 6119889119906) 09427 08237 09288

1198652(6 times 6119889V) 09370 08241 09283

1198653(8 times 8) 09430 08274 09312

1198654(12 times 12) 09605 08331 09505

1198655(17 times 17) 09601 08495 09746

classifiedTheworkflow of Strategy 1 is shown in Figure 4 anddescribed by the following equation

119891 (C119899+119897) = sgn(1198772 minus

119899

sum

119894119895=1

120572119894120572119895120581 (C119894C119895)

+ 2sum

119894

120572119894120581 (C119894C119899+119897) minus 120581 (C

119899+119897C119899+119897))

= 1 119891 (C

119899+119897) ge 0

minus1 119891 (C119899+119897) lt 0

(24)

whereC119899+119897

is the covariance matrix descriptor of the (119899+ 119897)thframe needed to be classified and C

119894and C

119895are the samples

of the dictionary CD ldquo1rdquo corresponds to the normal frameldquominus1rdquo corresponds to the abnormal frame

42 Strategy 2 In this strategy the dictionary is updatedthrough both training and testing periodsThe feature extrac-tion step (Step 1) and the online training steps (Steps 2 and 3)are the same as the ones presented in Strategy 1 The testingstep is different The new coming datum which is detectedas normal but satisfies dictionary update condition shouldbe included into CD The dictionary is needed to be updatedthrough the testing period to include new samples

Step 4 Strategy 2 If the incoming frame sample C119899+119897

isclassified as normal (119891(C

119899+119897) = 1) the data is checked by

the criterion described in Section 3 When the data satisfies

Journal of Electrical and Computer Engineering 7

(a) Normal lawn scene (b) Abnormal lawn scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC lawn offline

094092

09088086084082

080 005 01 015 02

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

094092

09088086084082

080 005 01 015 02

ROC lawn Strategy 1

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 5 Detection results of the lawn scene (a) The detection result of one normal frame (b) The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the lawn scene results via one-class SVM All the training samples are learned together offlineThe biggest AUC value is 09591 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUC value is09581

the dictionary update criterion this testing sample will beincluded into the dictionary This step can be generalized bythe following equation

119891 (C119899+119897) = sgn(1198772 minus

119899

sum

119894119895=1

120572119894120572119895120581 (C119894C119895)

+ 2sum

119894

120572119894120581 (C119894C119899+119897) minus 120581 (C

119899+119897C119899+119897))

=

1 119891 (C119899+119897) ge 0

120598119905ge 1205830997888rarr CD = CD cup C

119899+119897

120598119905lt 1205830997888rarr CD = CD

minus1 119891 (C119899+119897) lt 0

(25)

5 Abnormal Detection Result

This section presents the results of experiments conducted toanalyze the performance of the proposedmethod A compet-itive performance through both Strategy 1 and Strategy 2 ofUMN [31] dataset is presented

51 Abnormal Visual Events Detection Strategy 1 The resultsof the proposed abnormal events detection method viaStrategy 1 online one-class SVM of UMN [31] dataset areshown below

The UMN dataset includes eleven video sequences ofthree different scenes (the lawn indoor and plaza) ofcrowded escape events The normal samples for training or

8 Journal of Electrical and Computer Engineering

(a) Normal indoor scene (b) Abnormal indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC indoor offline

085

08

075

07

065

0 01 02 03 04

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC indoor Strategy 1

085

08

075

07

065

0 01 02 03 04

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 6 Detection results of the indoor scene (a)The detection result of one normal frame (b)The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the indoor scene results via one-class SVM All the training samples are learned togetheroffline The biggest AUC value is 08649 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUCvalue is 08628

Table 3 The comparison of our proposed method with the state-of-the-art methods for the abnormal event detection in the UMNdataset

Method Area under ROCLawn Indoor Plaza

Social force [16] 096Optical flow [16] 084NN [17] 093SRC [17] 0995 0975 0964STCOG [18] 09362 07759 09661COV online (ours) 09605 08628 09746

for normal testing are the frames where the people are walk-ing in different directions The samples for abnormal testingare the frames where people are running The detectionresults of the lawn scene indoor scene and plaza scene areshown in Figures 5 6 and 7 respectively Gaussian kernelfor the Lie Group is used in these three scenes Differentvalues of120590 and penalty factor119862 are chosen the area under theROC curve is shown as a function of these parameters [32]The results show that taking covariance matrix as descriptorcan obtain satisfactory performance for abnormal detectionAnd also training the samples online can obtain similarlydetection performance as training all the samples offline

Journal of Electrical and Computer Engineering 9

(a) Normal plaza scene (b) Abnormal plaza scene

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

ROC plaza offline

False positive

True

pos

itive

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

False positive

True

pos

itive

ROC plaza Strategy 1

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 7 Detection results of the plaza scene (a) The detection result of one normal frame (b) The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the plaza scene results via one-class SVM All the training samples are learned together offlineThe biggest AUC value is 09649 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUC value is09632

Online one-class SVM is appropriate to detect abnormalvisual events There are 1431 frames in the lawn scene and480 normal frames are used for training In the offline strat-egy all the 480 frames covariancematrices should be saved inthememory In Strategy 1 100 frames covariancematrices areconsidered as the dictionary firstlyWhen119865

5-17times17 feature is

adopted to construct the covariance descriptor the varianceof Gaussian kernel is 120590 = 1 the preset threshold of thecriterion is 120583

0= 05 the dictionary size increases from 100

to 101 and the maximum accuracy of the detection resultsis 9169 In the indoor scene there are 2975 normal framesand 1057 abnormal frames In the plaza scene there are 1831normal frames and 286 abnormal framesTheprocesses of theexperiments are similar to the ones of the lawn scene Whenfeature vector is 119865

5-17 times 17 120590 = 1 120583

0= 05 the dictionary

size of these two scenes remain 100The online strategy keepsthe memory size almost unchanged when the size of trainingdataset increases

52 Abnormal Visual Events Detection Strategy 2 The resultsof the abnormal event detection method via Strategy 2 ofUMNdataset are shown as follows In the experiment processof the lawn scene 100 normal samples from the trainingsamples are learnt firstly and then the other 380 trainingdata are learnt online one by one After these two trainingsteps we can obtain the basic dictionary from the trainingsamples and also the classifier In the following testing stepthe dictionary is updated if the sample satisfies the dictionaryupdate criterion When a new sample is coming it is firstlydetected by the previous classifier If it is classified as anomaly

10 Journal of Electrical and Computer Engineering

094092

09088086084082

08

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC lawn Strategy 2

F1

F2

F3

F4

F5

(a) ROC Strategy 2 lawn scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

085

08

075

07

065

0 01 02 03 04

ROC indoor Strategy 2

F1

F2

F3

F4

F5

(b) ROC Strategy 2 indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC plaza Strategy 2

F1

F2

F3

F4

F5

098096094092

09088086084082

08

(c) ROC Strategy 2 plaza scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC train data learnt offline

LawnPlazaIndoor

(d) ROC train offline

Figure 8 ROC curve of UMN dataset (a) ROC curve of different features 119865 results via Strategy 2 of the lawn sceneThe biggest AUC value is09605 (b) Strategy 2 results of the indoor scene The biggest AUC value is 08495 (c) Strategy 2 results of the plaza scene The biggest AUCvalue is 09746 (d)The ROC curve of best performance of the lawn indoor and plaza scene when the training samples are learnt offlineThebiggest AUC values of the lawn indoor and plaza are 09591 08649 and 09649

the dictionary and the classifier are not changed Otherwiseif the sample is classified as a normal one the sparse criterionintroduced in Section 3 is used to check the correlationbetween the earlier dictionary and this new datum It willbe included into the dictionary when it satisfied the updatecondition The dictionary will be updated through the wholetesting period The other two scenes the indoor and plazascene are handled by the same methods When 119865

5-17 times 17

feature is adopted the variance of the Gaussian kernel is120590 = 1 and the preset threshold of the criterion is 120583

0= 05

and the dictionary size of the lawn indoor and plaza scene

is increased from 100 to 106 102 and 102 respectively TheROC curve of detection results of these three scenes is shownin Figures 8(a) 8(b) and 8(c) Besides the merit of savingmemory of Strategy 1 Strategy 2 also has the advantage ofadaptation to the long duration sequence

The results performances of offline strategy Strategy 1and Strategy 2 are shown in Table 2 The performances ofthese two strategies results are similar to that of the resultswhen all training samples are learnt together When 119865

4(12 times

12) or1198655(17times17) are chosen as the features to formcovariance

matrix descriptor the results have the best performance

Journal of Electrical and Computer Engineering 11

These two features are more abundant to include movementand intensity information

The result performances of the covariancematrix descrip-tor based online one-class SVM method and the state-of-the-art methods are shown in Table 3 The covariance matrixbased online abnormal frame detection method obtainscompetitive performance In general our method is betterthan others except sparse reconstruction cost (SRC) [17]in lawn scene and indoor scene In that paper multiscaleHOF is taken as a feature and a testing sample is classi-fied by its sparse reconstructor cost through a weightedlinear reconstruction of the overcomplete normal basis setBut computation of the HOF might takes more time thancalculating covariance By adopting the integral image [15]the covariance matrix descriptor of the subimage can becomputed conveniently So the covariance descriptor can beappropriately used to analyze the partial movement

6 Conclusions

A method for the abnormal event detection of the frameis proposed The method consists of covariance matrixdescriptor encoding the movement features and the onlinenonlinear one-class SVM classification method We havedeveloped two nonlinear one-class SVM based abnormalevent detection techniques that update the normal modelsof the surveillance video data in an online framework Theproposed algorithm has been tested on a video datasetyielding successful results to detect abnormal events

Acknowledgments

This work is partially supported by China ScholarshipCouncil of Chinese Government SURECAP CPER Projectfunded by Region Champagne-Ardenne and CAPSECCRCA FEDER platform

References

[1] F Jiang J Yuan S A Tsaftaris and A K Katsaggelos ldquoAnoma-lous video event detection using spatiotemporal contextrdquo Com-puter Vision and Image Understanding vol 115 no 3 pp 323ndash333 2011

[2] C Piciarelli C Micheloni and G L Foresti ldquoTrajectory-basedanomalous event detectionrdquo IEEE Transactions on Circuits andSystems for Video Technology vol 18 no 11 pp 1544ndash1554 2008

[3] C Piciarelli and G L Foresti ldquoOn-line trajectory clustering foranomalous events detectionrdquo Pattern Recognition Letters vol27 no 15 pp 1835ndash1842 2006

[4] T Xiang and S Gong ldquoIncremental and adaptive abnormalbehaviour detectionrdquo Computer Vision and Image Understand-ing vol 111 no 1 pp 59ndash73 2008

[5] S Gong and T Xiang ldquoRecognition of group activities usingdynamic probabilistic networksrdquo in Proceedings of the 9th IEEEInternational Conference on Computer Vision (ICCV rsquo03) pp742ndash749 October 2003

[6] J KimandKGrauman ldquoObserve locally infer globally a space-time MRF for detecting abnormal activities with incrementalupdatesrdquo in Proceedings of the IEEE Conference on Computer

Vision and Pattern Recognition (CVPR rsquo09) pp 2921ndash2928Miami Fla USA June 2009

[7] A Adam E Rivlin I Shimshoni and D Reinitz ldquoRobustreal-time unusual event detection using multiple fixed-locationmonitorsrdquo IEEE Transactions on Pattern Analysis and MachineIntelligence vol 30 no 3 pp 555ndash560 2008

[8] O Boiman andM Irani ldquoDetecting irregularities in images andin videordquo International Journal of Computer Vision vol 74 no1 pp 17ndash31 2007

[9] X Zhu Z Liu and J Zhang ldquoHuman activity clustering foronline anomaly detectionrdquo Journal of Computers vol 6 no 6pp 1071ndash1079 2011

[10] X Zhu and Z Liu ldquoHuman behavior clustering for anomalydetectionrdquo Frontiers of Computer Science in China vol 5 no3 pp 279ndash289 2011

[11] Y Benezeth P M Jodoin and V Saligrama ldquoAbnormalitydetection using low-level co-occurring eventsrdquo Pattern Recog-nition Letters vol 32 no 3 pp 423ndash431 2011

[12] Y Benezeth P M Jodoin V Saligrama and C RosenbergerldquoAbnormal events detection based on spatio-temporal co-occurencesrdquo in Proceedings of the IEEE Conference on ComputerVision and Pattern Recognition (CVPR rsquo09) pp 2458ndash2465Miami Fla USA June 2009

[13] A Mittal A Monnet and N Paragios ldquoScene modeling andchange detection in dynamic scenes a subspace approachrdquoComputer Vision and Image Understanding vol 113 no 1 pp63ndash79 2009

[14] B K P Horn and B G Schunck ldquoDetermining optical flowrdquoArtificial Intelligence vol 17 no 1ndash3 pp 185ndash203 1981

[15] O Tuzel F Porikli and P Meer ldquoRegion covariance afast descriptor for detection and classificationrdquo in ComputerVisionmdashECCV 2006 vol 3952 of Lecture Notes in ComputerScience pp 589ndash600 Springer New York NY USA 2006

[16] RMehran A Oyama andM Shah ldquoAbnormal crowd behaviordetection using social force modelrdquo in Proceedings of the IEEEConference on Computer Vision and Pattern Recognition (CVPRrsquo09) pp 935ndash942 Miami Fla USA June 2009

[17] Y Cong J Yuan and J Liu ldquoSparse reconstruction cost forabnormal event detectionrdquo in Proceedings of the IEEE Confer-ence on Computer Vision and Pattern Recognition (CVPR rsquo11)pp 3449ndash3456 Providence RI USA June 2011

[18] Y Shi Y Gao and R Wang ldquoReal-time abnormal eventdetection in complicated scenesrdquo in Proceedings of the 20thInternational Conference on Pattern Recognition (ICPR rsquo10) pp3653ndash3656 Istanbul Turkey August 2010

[19] B Scholkopf and A J Smola Learning with Kernels SupportVector Machines Regularization Optimization and BeyondMIT Press Boston Mass USA 2002

[20] B Hall Lie Groups Lie Algebras And Representations AnElementary Introduction vol 222 Springer New York NYUSA 2003

[21] C Gao F Li and C Shen ldquoResearch on lie group kernellearning algorithmrdquo Journal of Frontiers of Compter Science andTechnology vol 6 pp 1026ndash1038 2012

[22] V N Vapnik and A Lerner ldquoPattern recognition using general-ized portrait methodrdquo Automation and Remote Control vol 24pp 774ndash780 1963

[23] V Vapnik The Nature of Statistical Learning Theory SpringerNew York NY USA 2000

[24] D TaxOne-class classification [PhD thesis] Delft University ofTechnology 2001

12 Journal of Electrical and Computer Engineering

[25] D M Tax and R P Duin ldquoData domain description usingsupport vectorsrdquo in Proceedings of the European Symposiumon Artificial Neural Networks Computational Intelligence andMachine Learning (ESANN rsquo99) vol 99 pp 251ndash256 1999

[26] B Scholkopf J C Platt J Shawe-Taylor A J Smola and RC Williamson ldquoEstimating the support of a high-dimensionaldistributionrdquo Neural Computation vol 13 no 7 pp 1443ndash14712001

[27] Z Noumir P Honeine and C Richard ldquoOnline one-classmachines based on the coherence criterionrdquo in Proceedings ofthe 20th European Signal Processing Conference (EUSIPCO rsquo12)pp 664ndash668 2012

[28] Z Noumir P Honeine and C Richard ldquoOne-class machinesbased on the coherence criterionrdquo in Proceedings of the IEEEStatistical Signal Processing Workshop (SSP rsquo12) pp 600ndash6032012

[29] P Honeine ldquoOnline kernel principal component analysis areducedorder modelrdquo IEEE Transactions on Pattern Analysisand Machine Intelligence vol 34 no 9 pp 1814ndash1826 2012

[30] C Richard J C M Bermudez and P Honeine ldquoOnlineprediction of time series data with kernelsrdquo IEEE Transactionson Signal Processing vol 57 no 3 pp 1058ndash1067 2009

[31] UMN ldquoUnusual crowd activity dataset of university of Min-nesotardquo 2006 httpmhacsumneduproj eventsshtml

[32] J A Hanley and B J McNeil ldquoThe meaning and use of thearea under a receiver operating characteristic (ROC) curverdquoRadiology vol 143 no 1 pp 29ndash36 1982

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

2 Journal of Electrical and Computer Engineering

support vector machine algorithm is applied in an onlinefashion in order to detect abnormal events The featuresare extracted based on the optical flow which presents themovement information Experiments of real surveillancevideo dataset show that our online abnormal detection tech-niques can obtain satisfactory performance The rest of thepaper is organized as follows In Section 2 covariance matrixdescriptor of motion feature is introduced In Section 3 theonline one-class SVM classification method is presented InSection 4 two abnormal detection strategies based on onlinenonlinear one-class SVM are proposed In Section 5 wepresent results of real-world video scenes Finally Section 6concludes the paper

2 Covariance Descriptor of Frame Behavior

The optical flow is a feature which presents the directionand the amplitude of a movement It can provide importantinformation about the spatial arrangement of the objects andthe change rate of this arrangement [14] We adopt Horn-Schunck (HS) optical flow computation method in our workThe optical flow of the gray scale image is formulated as theminimizer of the following global energy functional

119864 = ∬[(119868119909119906 + 119868119910V + 119868119905)2

+ 1205722

(nabla1199062

+ nablaV2)] 119889119909119889119910 (1)

where 119868 is the intensity of the image 119868119909 119868119910 and 119868

119905are the

derivatives of the image intensity value along the 119909 119910 andtime 119905 dimension 119906 and V are the components of the opticalflow in the horizontal and vertical direction and 120572 representsthe weight of the regularization term

We introduce the covariance matrix encoding the opticalflow and intensity of each frame as the descriptor to representthemovementThe covariance feature descriptor is originallyproposed by Tuzel et al [15] for pattern matching in a targettracking problem The descriptor is defined as

119865 (119909 119910 119894) = 120601119894(119868 119909 119910) (2)

where 119868 is the color information of an image (which can begray RGB HSV HLS etc) 120601

119894is a mapping relating the image

with the 119894th feature from the image 119865 is the 119882 times 119867 times 119889

dimensional feature extracted from image 119868119882 and119867 are theimage width and image height and 119889 is the number of chosenfeatures For each frame the feature can be represented as119889 times 119889 covariance matrix

C =1

119899 minus 1

119899

sum

119896=1

(z119896minus 120583) (z

119896minus 120583)⊤

(3)

where 119899 is the number of the pixels sampled in the framez119896is the feature vector of pixel 119896 120583 is the mean of all the

selected points and C is the covariance matrix of the featurevector 119865 The covariance descriptor C of each frame dose nothave any information regarding the sample ordering and thenumber of points [15] Because the feature 119865 can be designedas different approaches to fuse features the covariancematrixdescriptor proposes a way to merge multiple parametersDifferent choices of feature vectors extraction are shown in

Table 1 Different choices of feature 119865 to construct the covariancedescriptor

Feature vector 1198651198651(6 times 6) [119910 119909 119906 V 119906

119909119906119910]

1198652(6 times 6) [119910 119909 119906 V V

119909V119910]

1198653(8 times 8) [119910 119909 119906 V 119906

119909119906119910V119909V119910]

1198654(12 times 12) [119910 119909 119906 V 119906

119909119906119910V119909V119910119906119909119909

119906119910119910

V119909119909

V119910119910]

1198655(17 times 17) [119910 119909 119906 V 119906

119909119906119910V119909V119910119906119909119909

119906119910119910

V119909119909

V119910119910

119868 119868119909119868119910119868119909119909

119868119910119910]

Table 1 where 119868 is the intensity of the gray image 119906 and Vare horizontal and vertical components of optical flow 119868

119909

119906119909 and V

119909are the first derivatives of the intensity horizontal

optical flow and vertical optical flow in the 119909 directionrespectively 119868

119910 119906119910 and V

119910are the first derivatives of the

corresponding feature in the 119910 direction and respectively119868119909119909 119906119909119909 and V

119909119909are the second derivatives in 119909 direction

119868119910119910 119906119910119910 and V

119910119910are the second derivatives in 119910 direction

The flowchart of covariance matrix descriptor computationis shown in Figure 2 The optical flow and correspondingpartial derivative characterize the interframe information orcan be regarded as the movement information The intensityof the frame and partial derivative of the intensity describethe intraframe information they encode the appearanceinformation of the frame

If proper parameters are given the traditionally usedkernel such as Gaussian polynomial and sigmoidal kernelhas similar performances [19] Gaussian kernel 120581(x

119894 x119895) =

exp(minusx119894minus x1198952

21205902

) is chosen for our spatial features Thecovariance matrix is an element in Lie group the Gaussiankernel on the Euclidean spaces is not suitable for the covari-ance descriptorsTheGaussian kernel in Lie Group is definedas [20 21]

120581 (X119894X119895) = exp(minus

10038171003817100381710038171003817log (Xminus1

119894X119895)10038171003817100381710038171003817

21205902) (X

119894X119895) isin 119866 times 119866

(4)

where X119894and X

119895are matrices in Lie Group 119866

3 Online One-Class SVM

The essence of an abnormal detection problem is that onlynormal scene samples are available The one-class SVMframework is well suitable to an abnormal detection problemSupport vector machine (SVM) is initially proposed byVapnik and Lerner [22 23] It is a method based on statisticallearning theory and has fine performance to classify data andrecognize patterns There are two frameworks of one-classSVM one is support vector data description (SVDD) whichis presented in [24 25] and the other is ]-support vectorclassifier (]-SVC) introduced in [26]The SVDD formulationis adopted in our work It computes a sphere shaped decisionboundary with minimal volume around a set of objects The

Journal of Electrical and Computer Engineering 3

(a) Normal lawn scene (b) Abnormal lawn scene

Figure 1 Examples of the normal and abnormal scenes (a) Normal lawn scene all the people are walking (b) Abnormal lawn scene all thepeople are running

Consecutive frameFramei

Framei+1

Optical flow field OPi

Features

F(x y j)

j = 1 2 nCframe119894

Figure 2 Covariance descriptor computation of a video frame

center of the sphere c and the radius 119877 are to be determinedvia the following optimization problem

min119877120585119888

1198772

+ 119862

119899

sum

119894=1

120585119894 (5)

subject to 1003817100381710038171003817Φ (x119894) minus c1003817100381710038171003817

2

le 1198772

+ 120585119894 120585119894ge 0 forall119894 (6)

where 119899 is the number of training samples and 120585119894is the

slack variable for penalizing the outliersThe hyperparameter119862 is the weight for restraining slack variables it tunesthe number of acceptable outliers The nonlinear functionΦ X rarr H maps a datum x

119894into the feature space

H it allows to solve a nonlinear classification problem bydesigning a linear classifier in the feature space H 120581 is thekernel function for computing dot products inH 120581(x x1015840) =⟨Φ(x) Φ(x1015840)⟩ By introducing Lagrange multipliers the dual

problem associated with (6) is written by the followingquadratic optimization problem

max120572

119899

sum

119894=1

120572119894120581 (x119894 x119894) minus

119899

sum

119894119895=1

120572119894120572119895120581 (x119894 x119895)

subject to 0 le 120572119894le 119862

119899

sum

119894=1

120572119894= 1

c =119899

sum

119894=1

120572119894Φ(x119894)

(7)

The decision function is

119891 (x) = sgn(1198772 minus119899

sum

119894119895=1

120572119894120572119895120581 (x119894 x119895)

+ 2

119899

sum

119894=1

120572119894120581 (x119894 x) minus 120581 (x x))

(8)

For the large training data the solution cannot beobtained easily and an online strategy to train the data is

4 Journal of Electrical and Computer Engineering

used in our work Let cD denotes a sparse model of the centerc119899= (1119899)sum

119899

119894=1Φ(x119894) by using a small subset of available

samples which is called dictionary

cD = sum

119894isinD

120572119894Φ(x119894) (9)

whereD sub 1 2 119899 and let119873D denote the cardinality ofthis subset xD The distance of a mapped datum Φ(x) withrespect to the center cD can be calculated by

1003817100381710038171003817Φ (x) minus cD1003817100381710038171003817 = sum

119894119895isinD

120572119894120572119895120581 (x119894 x119895)

minus 2sum

119894isinD

120572119894120581 (x119894 x) + 120581 (x x)

(10)

A modification of the original formulation of the one-class classification algorithm that consists of minimizing theapproximation error c

119899minus cD is [27 28]

120572 = arg min120572119894119894isinD

1003817100381710038171003817100381710038171003817100381710038171003817

1

119899

119899

sum

119894=1

Φ(x119894) minus sum

119894isinD

120572119894Φ(x119894)

1003817100381710038171003817100381710038171003817100381710038171003817

2

(11)

The final solution is given by

120572 = Kminus1120581 (12)

where K is the Grammatrix with (119894 119895)th entry 120581(x119894 x119895) and 120581

is the column vector with entries (1119899)sum119899119894=1

120581(x119896 x119894) 119896 isin D

In the online scheme at each time step there is a newsample Let 120572

119899denote the coefficients K

119899denote the Gram

matrix and 120581119899denote the vector at time step 119899 A criterion is

used to determine whether the new sample can be includedinto the dictionary A threshold 120583

0is preset for the datum

x119905at time step 119905 the coherence-based sparsification criterion

[29 30] is

120598119905= max119894isinD

1003816100381610038161003816120581 (x119894 x119905)1003816100381610038161003816 (13)

First Case (120598119905gt 1205830) In this case the new data Φ(x

119899+1) is not

included into the dictionaryThe GrammatrixK119899+1

= K119899 120581119899

changes online

120581119899+1

=1

119899 + 1(119899120581119899+ b)

120572119899+1

= Kminus1119899+1120581119899+1

=119899

119899 + 1120572119899+

1

119899 + 1Kminus1119899b

(14)

where b is the column vector with entries 120581(x119894 x119899+1

) 119894 isin D

Second Case (120598119905le 1205830) In this case the new data Φ(x

119899+1) is

included into the dictionaryD The Grammatrix K changes

K119899+1

= [

[

K119899

bb⊤ 120581 (x

119899+1 x119899+1

)]

]

(15)

By using Woodbury matrix identity

(119860 + 119880119862119881)minus1

= 119860minus1

minus 119860minus1

119880(119862minus1

+ 119881119860minus1

119880)minus1

119881119860minus1

(16)

Kminus1119899+1

can be calculated iteratively

Kminus1119899+1

= [

[

Kminus1119899

00⊤ 0

]

]

+1

120581 (x119899+1

x119899+1

) minus b⊤Kminus1119899b

times [

[

minusKminus1119899b

1]

]

[minusb⊤Kminus11198991]

(17)

The vector 120581119899+1

is updated from 120581119899

120581119899+1

=1

119899 + 1[119899120581119899+

120581119899+1

] (18)

with

120581119899+1

=

119899+1

sum

119894=1

120581 (x119899+1

x119894) (19)

Computing 120581119899+1

as (19) needs to save all the samples x119899+1119894=1

inmemory For conquering this issue it can compute as 120581

119899+1=

(119899 + 1)120581(x119899+1

x119899+1

) by considering an instant estimationTheupdate of 120572

119899+1from 120572

119899is

120572119899+1

=1

119899 + 1[119899120572119899+ Kminus1119899b

0]

minus1

(119899 + 1) (120581 (x119899+1

x119899+1

) minus b⊤Kminus1119899b)

times [

[

Kminus1119899b

1]

]

(119899b⊤120572119899+ b⊤Kminus1

119899b minus 120581119899+1

)

(20)

Based on (20) we have an online implementation of the one-class SVM learning phase

4 Abnormal Events Detection

In an abnormal event detection problem it is assumedthat a set of training frames 119868

1 119868

119899 (the positive class)

describing the normal behavior is obtained The generalarchitectures of abnormal detection are introduced below

The offline training strategy refers to the case where allthe training samples are learnt as one batch as shown inFigure 3(a) We propose two abnormal detection strategiesthe difference between these two strategies is the time whenthe dictionary is fixed These two strategies are shown inFigures 3(b) and 3(c) Strategy 1 is shown in Figure 3(b) Thetraining data are learnt one by oneWhen the training periodis finished the dictionary and the classifier are fixed Eachtest datum is classified based on the dictionary Figure 3(c)illustrates Strategy 2 The training procedure is the same asStrategy 1 But in the testing period the dictionary is updatedif the datum x

119894satisfies the dictionary update condition The

details of these two strategies are explained in the following

Journal of Electrical and Computer Engineering 5

Train offline

Test onlinem n minus m

(a) Strategy offline

Test online

Train online

m n minus m

Train

Dictionary fixed

(b) Strategy 1

Train

Dictionary fixed

Train and test online

Test onlinem n minus m

(c) Strategy 2

Figure 3 Offline and two online abnormal event detection strategies based on one-class SVM (a) Strategy offlineThe training data are learntas one batch offline (b) Strategy 1 The dictionary is fixed when all the training data are learnt (c) Strategy 2 The dictionary continues beingupdated through the testing period

Features selectionon original image

Learning step (online)Covariance

Covariance

optical flow

Classification

SVM train

(online training)

people walk

Detection step (online)One-class SVM

R

Origin

Abnormal eventdetection

people run

optical flow

x11 x12 middot middot middot

middot middot middot

middot middot middot

x1n

x21 x22 x2n

xn1 xn2 xnn

x11 x12 middot middot middot

middot middot middot

middot middot middot

x1n

x21 x22 x2n

xn1 xn2 xnn

⋱

⋱

Figure 4 Major processing states of the proposed abnormal frame event detection method The covariance of the frame is computed

6 Journal of Electrical and Computer Engineering

41 Strategy 1 In Strategy 1 the dictionary is updated merelythrough the training period

Step 1 The first step is calculating the covariance matrixdescriptor of training frames based on the image intensity andthe optical flow This step can be generalized as

(1198681OP1) (1198682OP2) (119868

119899OP119899) 997888rarr C

1C2 C

119899

(21)

where (I1OP1) (I2OP2) (119868

119899OP119899) are the image

intensity and the corresponding optical flow of the 1st to 119899thframe C

1C2 C

119899 are the covariance matrix descriptors

Step 2 The second step consists of applying one-class SVMon the small subset of extracted descriptor of the trainingnormal frames to obtain the support vectors Consider asubset C

119894119898

119894=1 1 le 119898 ≪ 119899 of data selected from the full

training sample set C119894119899

119894=1 without loss of generality and

assume that the first 119898 examples are chosen This set of 119898examples is called dictionary CD

C1C2 C

119898 1 le 119898 ≪ 119899

SVM997888rarr support vector 119878119901

1 1198781199012 119878119901

119900

(22)

where the set C1C2 C

119898 is the first119898 covariancematrix

descriptors of the training frames it is the original dictionaryCD In one-class SVM the majority of the training samplesdo not contribute to the definition of the decision functionThe entries of a monitory subset of the training samples1198781199011 1198781199012 119878119901

119900 119900 le 119898 are support vectors contributing

to the definition of the decision function

Step 3 After learning the dictionary CD which includesthe first 119898 1 le 119898 ≪ 119899 samples the training samplesC119898+1

C119898+2

C119899 are learned online via the technique

described in Section 3 This step can be generalized as

CDC119896 119898 lt 119896 le 119899SVM997888rarr support vector 119878119901

1 1198781199012 119878119901

119901

119900 le 119901 le 119899

CD = CD cup C119896 if 120598

119905ge 1205830

(23)

where CD is the dictionary obtained through Step 2 C119896is a

new sample in the remaining training dataset According tothe criterion introduced in Section 3 if the new sample C

119896

satisfies the dictionary updated condition it will be includedinto the dictionary CD

Step 4 Based on the dictionary and the classifier obtainedfrom the training frames the incoming frame sample C

119899+119897is

Table 2 AUC of abnormal event detection method of differentfeatures 119865 via original SVDD which learns training samples offlineStrategy 1 online one-class SVM and Strategy 2 online one-classSVMThe biggest value of each method is shown in bold

Features Area under ROCLawn Indoor Plaza

Training samples are learned offline1198651(6 times 6119889119906) 09426 08351 09323

1198652(6 times 6119889V) 09400 08358 09321

1198653(8 times 8) 09440 08375 09359

1198654(12 times 12) 09591 08440 09580

1198655(17 times 17) 09567 08649 09649

Strategy 11198651(6 times 6119889119906) 09399 08328 09343

1198652(6 times 6119889V) 09390 08355 09366

1198653(8 times 8) 09418 08377 09411

1198654(12 times 12) 09581 08457 09573

1198655(17 times 17) 09551 08628 09632

Strategy 21198651(6 times 6119889119906) 09427 08237 09288

1198652(6 times 6119889V) 09370 08241 09283

1198653(8 times 8) 09430 08274 09312

1198654(12 times 12) 09605 08331 09505

1198655(17 times 17) 09601 08495 09746

classifiedTheworkflow of Strategy 1 is shown in Figure 4 anddescribed by the following equation

119891 (C119899+119897) = sgn(1198772 minus

119899

sum

119894119895=1

120572119894120572119895120581 (C119894C119895)

+ 2sum

119894

120572119894120581 (C119894C119899+119897) minus 120581 (C

119899+119897C119899+119897))

= 1 119891 (C

119899+119897) ge 0

minus1 119891 (C119899+119897) lt 0

(24)

whereC119899+119897

is the covariance matrix descriptor of the (119899+ 119897)thframe needed to be classified and C

119894and C

119895are the samples

of the dictionary CD ldquo1rdquo corresponds to the normal frameldquominus1rdquo corresponds to the abnormal frame

42 Strategy 2 In this strategy the dictionary is updatedthrough both training and testing periodsThe feature extrac-tion step (Step 1) and the online training steps (Steps 2 and 3)are the same as the ones presented in Strategy 1 The testingstep is different The new coming datum which is detectedas normal but satisfies dictionary update condition shouldbe included into CD The dictionary is needed to be updatedthrough the testing period to include new samples

Step 4 Strategy 2 If the incoming frame sample C119899+119897

isclassified as normal (119891(C

119899+119897) = 1) the data is checked by

the criterion described in Section 3 When the data satisfies

Journal of Electrical and Computer Engineering 7

(a) Normal lawn scene (b) Abnormal lawn scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC lawn offline

094092

09088086084082

080 005 01 015 02

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

094092

09088086084082

080 005 01 015 02

ROC lawn Strategy 1

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 5 Detection results of the lawn scene (a) The detection result of one normal frame (b) The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the lawn scene results via one-class SVM All the training samples are learned together offlineThe biggest AUC value is 09591 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUC value is09581

the dictionary update criterion this testing sample will beincluded into the dictionary This step can be generalized bythe following equation

119891 (C119899+119897) = sgn(1198772 minus

119899

sum

119894119895=1

120572119894120572119895120581 (C119894C119895)

+ 2sum

119894

120572119894120581 (C119894C119899+119897) minus 120581 (C

119899+119897C119899+119897))

=

1 119891 (C119899+119897) ge 0

120598119905ge 1205830997888rarr CD = CD cup C

119899+119897

120598119905lt 1205830997888rarr CD = CD

minus1 119891 (C119899+119897) lt 0

(25)

5 Abnormal Detection Result

This section presents the results of experiments conducted toanalyze the performance of the proposedmethod A compet-itive performance through both Strategy 1 and Strategy 2 ofUMN [31] dataset is presented

51 Abnormal Visual Events Detection Strategy 1 The resultsof the proposed abnormal events detection method viaStrategy 1 online one-class SVM of UMN [31] dataset areshown below

The UMN dataset includes eleven video sequences ofthree different scenes (the lawn indoor and plaza) ofcrowded escape events The normal samples for training or

8 Journal of Electrical and Computer Engineering

(a) Normal indoor scene (b) Abnormal indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC indoor offline

085

08

075

07

065

0 01 02 03 04

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC indoor Strategy 1

085

08

075

07

065

0 01 02 03 04

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 6 Detection results of the indoor scene (a)The detection result of one normal frame (b)The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the indoor scene results via one-class SVM All the training samples are learned togetheroffline The biggest AUC value is 08649 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUCvalue is 08628

Table 3 The comparison of our proposed method with the state-of-the-art methods for the abnormal event detection in the UMNdataset

Method Area under ROCLawn Indoor Plaza

Social force [16] 096Optical flow [16] 084NN [17] 093SRC [17] 0995 0975 0964STCOG [18] 09362 07759 09661COV online (ours) 09605 08628 09746

for normal testing are the frames where the people are walk-ing in different directions The samples for abnormal testingare the frames where people are running The detectionresults of the lawn scene indoor scene and plaza scene areshown in Figures 5 6 and 7 respectively Gaussian kernelfor the Lie Group is used in these three scenes Differentvalues of120590 and penalty factor119862 are chosen the area under theROC curve is shown as a function of these parameters [32]The results show that taking covariance matrix as descriptorcan obtain satisfactory performance for abnormal detectionAnd also training the samples online can obtain similarlydetection performance as training all the samples offline

Journal of Electrical and Computer Engineering 9

(a) Normal plaza scene (b) Abnormal plaza scene

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

ROC plaza offline

False positive

True

pos

itive

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

False positive

True

pos

itive

ROC plaza Strategy 1

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 7 Detection results of the plaza scene (a) The detection result of one normal frame (b) The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the plaza scene results via one-class SVM All the training samples are learned together offlineThe biggest AUC value is 09649 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUC value is09632

Online one-class SVM is appropriate to detect abnormalvisual events There are 1431 frames in the lawn scene and480 normal frames are used for training In the offline strat-egy all the 480 frames covariancematrices should be saved inthememory In Strategy 1 100 frames covariancematrices areconsidered as the dictionary firstlyWhen119865

5-17times17 feature is

adopted to construct the covariance descriptor the varianceof Gaussian kernel is 120590 = 1 the preset threshold of thecriterion is 120583

0= 05 the dictionary size increases from 100

to 101 and the maximum accuracy of the detection resultsis 9169 In the indoor scene there are 2975 normal framesand 1057 abnormal frames In the plaza scene there are 1831normal frames and 286 abnormal framesTheprocesses of theexperiments are similar to the ones of the lawn scene Whenfeature vector is 119865

5-17 times 17 120590 = 1 120583

0= 05 the dictionary

size of these two scenes remain 100The online strategy keepsthe memory size almost unchanged when the size of trainingdataset increases

52 Abnormal Visual Events Detection Strategy 2 The resultsof the abnormal event detection method via Strategy 2 ofUMNdataset are shown as follows In the experiment processof the lawn scene 100 normal samples from the trainingsamples are learnt firstly and then the other 380 trainingdata are learnt online one by one After these two trainingsteps we can obtain the basic dictionary from the trainingsamples and also the classifier In the following testing stepthe dictionary is updated if the sample satisfies the dictionaryupdate criterion When a new sample is coming it is firstlydetected by the previous classifier If it is classified as anomaly

10 Journal of Electrical and Computer Engineering

094092

09088086084082

08

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC lawn Strategy 2

F1

F2

F3

F4

F5

(a) ROC Strategy 2 lawn scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

085

08

075

07

065

0 01 02 03 04

ROC indoor Strategy 2

F1

F2

F3

F4

F5

(b) ROC Strategy 2 indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC plaza Strategy 2

F1

F2

F3

F4

F5

098096094092

09088086084082

08

(c) ROC Strategy 2 plaza scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC train data learnt offline

LawnPlazaIndoor

(d) ROC train offline

Figure 8 ROC curve of UMN dataset (a) ROC curve of different features 119865 results via Strategy 2 of the lawn sceneThe biggest AUC value is09605 (b) Strategy 2 results of the indoor scene The biggest AUC value is 08495 (c) Strategy 2 results of the plaza scene The biggest AUCvalue is 09746 (d)The ROC curve of best performance of the lawn indoor and plaza scene when the training samples are learnt offlineThebiggest AUC values of the lawn indoor and plaza are 09591 08649 and 09649

the dictionary and the classifier are not changed Otherwiseif the sample is classified as a normal one the sparse criterionintroduced in Section 3 is used to check the correlationbetween the earlier dictionary and this new datum It willbe included into the dictionary when it satisfied the updatecondition The dictionary will be updated through the wholetesting period The other two scenes the indoor and plazascene are handled by the same methods When 119865

5-17 times 17

feature is adopted the variance of the Gaussian kernel is120590 = 1 and the preset threshold of the criterion is 120583

0= 05

and the dictionary size of the lawn indoor and plaza scene

is increased from 100 to 106 102 and 102 respectively TheROC curve of detection results of these three scenes is shownin Figures 8(a) 8(b) and 8(c) Besides the merit of savingmemory of Strategy 1 Strategy 2 also has the advantage ofadaptation to the long duration sequence

The results performances of offline strategy Strategy 1and Strategy 2 are shown in Table 2 The performances ofthese two strategies results are similar to that of the resultswhen all training samples are learnt together When 119865

4(12 times

12) or1198655(17times17) are chosen as the features to formcovariance

matrix descriptor the results have the best performance

Journal of Electrical and Computer Engineering 11

These two features are more abundant to include movementand intensity information

The result performances of the covariancematrix descrip-tor based online one-class SVM method and the state-of-the-art methods are shown in Table 3 The covariance matrixbased online abnormal frame detection method obtainscompetitive performance In general our method is betterthan others except sparse reconstruction cost (SRC) [17]in lawn scene and indoor scene In that paper multiscaleHOF is taken as a feature and a testing sample is classi-fied by its sparse reconstructor cost through a weightedlinear reconstruction of the overcomplete normal basis setBut computation of the HOF might takes more time thancalculating covariance By adopting the integral image [15]the covariance matrix descriptor of the subimage can becomputed conveniently So the covariance descriptor can beappropriately used to analyze the partial movement

6 Conclusions

A method for the abnormal event detection of the frameis proposed The method consists of covariance matrixdescriptor encoding the movement features and the onlinenonlinear one-class SVM classification method We havedeveloped two nonlinear one-class SVM based abnormalevent detection techniques that update the normal modelsof the surveillance video data in an online framework Theproposed algorithm has been tested on a video datasetyielding successful results to detect abnormal events

Acknowledgments

This work is partially supported by China ScholarshipCouncil of Chinese Government SURECAP CPER Projectfunded by Region Champagne-Ardenne and CAPSECCRCA FEDER platform

References

[1] F Jiang J Yuan S A Tsaftaris and A K Katsaggelos ldquoAnoma-lous video event detection using spatiotemporal contextrdquo Com-puter Vision and Image Understanding vol 115 no 3 pp 323ndash333 2011

[2] C Piciarelli C Micheloni and G L Foresti ldquoTrajectory-basedanomalous event detectionrdquo IEEE Transactions on Circuits andSystems for Video Technology vol 18 no 11 pp 1544ndash1554 2008

[3] C Piciarelli and G L Foresti ldquoOn-line trajectory clustering foranomalous events detectionrdquo Pattern Recognition Letters vol27 no 15 pp 1835ndash1842 2006

[4] T Xiang and S Gong ldquoIncremental and adaptive abnormalbehaviour detectionrdquo Computer Vision and Image Understand-ing vol 111 no 1 pp 59ndash73 2008

[5] S Gong and T Xiang ldquoRecognition of group activities usingdynamic probabilistic networksrdquo in Proceedings of the 9th IEEEInternational Conference on Computer Vision (ICCV rsquo03) pp742ndash749 October 2003

[6] J KimandKGrauman ldquoObserve locally infer globally a space-time MRF for detecting abnormal activities with incrementalupdatesrdquo in Proceedings of the IEEE Conference on Computer

Vision and Pattern Recognition (CVPR rsquo09) pp 2921ndash2928Miami Fla USA June 2009

[7] A Adam E Rivlin I Shimshoni and D Reinitz ldquoRobustreal-time unusual event detection using multiple fixed-locationmonitorsrdquo IEEE Transactions on Pattern Analysis and MachineIntelligence vol 30 no 3 pp 555ndash560 2008

[8] O Boiman andM Irani ldquoDetecting irregularities in images andin videordquo International Journal of Computer Vision vol 74 no1 pp 17ndash31 2007

[9] X Zhu Z Liu and J Zhang ldquoHuman activity clustering foronline anomaly detectionrdquo Journal of Computers vol 6 no 6pp 1071ndash1079 2011

[10] X Zhu and Z Liu ldquoHuman behavior clustering for anomalydetectionrdquo Frontiers of Computer Science in China vol 5 no3 pp 279ndash289 2011

[11] Y Benezeth P M Jodoin and V Saligrama ldquoAbnormalitydetection using low-level co-occurring eventsrdquo Pattern Recog-nition Letters vol 32 no 3 pp 423ndash431 2011

[12] Y Benezeth P M Jodoin V Saligrama and C RosenbergerldquoAbnormal events detection based on spatio-temporal co-occurencesrdquo in Proceedings of the IEEE Conference on ComputerVision and Pattern Recognition (CVPR rsquo09) pp 2458ndash2465Miami Fla USA June 2009

[13] A Mittal A Monnet and N Paragios ldquoScene modeling andchange detection in dynamic scenes a subspace approachrdquoComputer Vision and Image Understanding vol 113 no 1 pp63ndash79 2009

[14] B K P Horn and B G Schunck ldquoDetermining optical flowrdquoArtificial Intelligence vol 17 no 1ndash3 pp 185ndash203 1981

[15] O Tuzel F Porikli and P Meer ldquoRegion covariance afast descriptor for detection and classificationrdquo in ComputerVisionmdashECCV 2006 vol 3952 of Lecture Notes in ComputerScience pp 589ndash600 Springer New York NY USA 2006

[16] RMehran A Oyama andM Shah ldquoAbnormal crowd behaviordetection using social force modelrdquo in Proceedings of the IEEEConference on Computer Vision and Pattern Recognition (CVPRrsquo09) pp 935ndash942 Miami Fla USA June 2009

[17] Y Cong J Yuan and J Liu ldquoSparse reconstruction cost forabnormal event detectionrdquo in Proceedings of the IEEE Confer-ence on Computer Vision and Pattern Recognition (CVPR rsquo11)pp 3449ndash3456 Providence RI USA June 2011

[18] Y Shi Y Gao and R Wang ldquoReal-time abnormal eventdetection in complicated scenesrdquo in Proceedings of the 20thInternational Conference on Pattern Recognition (ICPR rsquo10) pp3653ndash3656 Istanbul Turkey August 2010

[19] B Scholkopf and A J Smola Learning with Kernels SupportVector Machines Regularization Optimization and BeyondMIT Press Boston Mass USA 2002

[20] B Hall Lie Groups Lie Algebras And Representations AnElementary Introduction vol 222 Springer New York NYUSA 2003

[21] C Gao F Li and C Shen ldquoResearch on lie group kernellearning algorithmrdquo Journal of Frontiers of Compter Science andTechnology vol 6 pp 1026ndash1038 2012

[22] V N Vapnik and A Lerner ldquoPattern recognition using general-ized portrait methodrdquo Automation and Remote Control vol 24pp 774ndash780 1963

[23] V Vapnik The Nature of Statistical Learning Theory SpringerNew York NY USA 2000

[24] D TaxOne-class classification [PhD thesis] Delft University ofTechnology 2001

12 Journal of Electrical and Computer Engineering

[25] D M Tax and R P Duin ldquoData domain description usingsupport vectorsrdquo in Proceedings of the European Symposiumon Artificial Neural Networks Computational Intelligence andMachine Learning (ESANN rsquo99) vol 99 pp 251ndash256 1999

[26] B Scholkopf J C Platt J Shawe-Taylor A J Smola and RC Williamson ldquoEstimating the support of a high-dimensionaldistributionrdquo Neural Computation vol 13 no 7 pp 1443ndash14712001

[27] Z Noumir P Honeine and C Richard ldquoOnline one-classmachines based on the coherence criterionrdquo in Proceedings ofthe 20th European Signal Processing Conference (EUSIPCO rsquo12)pp 664ndash668 2012

[28] Z Noumir P Honeine and C Richard ldquoOne-class machinesbased on the coherence criterionrdquo in Proceedings of the IEEEStatistical Signal Processing Workshop (SSP rsquo12) pp 600ndash6032012

[29] P Honeine ldquoOnline kernel principal component analysis areducedorder modelrdquo IEEE Transactions on Pattern Analysisand Machine Intelligence vol 34 no 9 pp 1814ndash1826 2012

[30] C Richard J C M Bermudez and P Honeine ldquoOnlineprediction of time series data with kernelsrdquo IEEE Transactionson Signal Processing vol 57 no 3 pp 1058ndash1067 2009

[31] UMN ldquoUnusual crowd activity dataset of university of Min-nesotardquo 2006 httpmhacsumneduproj eventsshtml

[32] J A Hanley and B J McNeil ldquoThe meaning and use of thearea under a receiver operating characteristic (ROC) curverdquoRadiology vol 143 no 1 pp 29ndash36 1982

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Journal of Electrical and Computer Engineering 3

(a) Normal lawn scene (b) Abnormal lawn scene

Figure 1 Examples of the normal and abnormal scenes (a) Normal lawn scene all the people are walking (b) Abnormal lawn scene all thepeople are running

Consecutive frameFramei

Framei+1

Optical flow field OPi

Features

F(x y j)

j = 1 2 nCframe119894

Figure 2 Covariance descriptor computation of a video frame

center of the sphere c and the radius 119877 are to be determinedvia the following optimization problem

min119877120585119888

1198772

+ 119862

119899

sum

119894=1

120585119894 (5)

subject to 1003817100381710038171003817Φ (x119894) minus c1003817100381710038171003817

2

le 1198772

+ 120585119894 120585119894ge 0 forall119894 (6)

where 119899 is the number of training samples and 120585119894is the

slack variable for penalizing the outliersThe hyperparameter119862 is the weight for restraining slack variables it tunesthe number of acceptable outliers The nonlinear functionΦ X rarr H maps a datum x

119894into the feature space

H it allows to solve a nonlinear classification problem bydesigning a linear classifier in the feature space H 120581 is thekernel function for computing dot products inH 120581(x x1015840) =⟨Φ(x) Φ(x1015840)⟩ By introducing Lagrange multipliers the dual

problem associated with (6) is written by the followingquadratic optimization problem

max120572

119899

sum

119894=1

120572119894120581 (x119894 x119894) minus

119899

sum

119894119895=1

120572119894120572119895120581 (x119894 x119895)

subject to 0 le 120572119894le 119862

119899

sum

119894=1

120572119894= 1

c =119899

sum

119894=1

120572119894Φ(x119894)

(7)

The decision function is

119891 (x) = sgn(1198772 minus119899

sum

119894119895=1

120572119894120572119895120581 (x119894 x119895)

+ 2

119899

sum

119894=1

120572119894120581 (x119894 x) minus 120581 (x x))

(8)

For the large training data the solution cannot beobtained easily and an online strategy to train the data is

4 Journal of Electrical and Computer Engineering

used in our work Let cD denotes a sparse model of the centerc119899= (1119899)sum

119899

119894=1Φ(x119894) by using a small subset of available

samples which is called dictionary

cD = sum

119894isinD

120572119894Φ(x119894) (9)

whereD sub 1 2 119899 and let119873D denote the cardinality ofthis subset xD The distance of a mapped datum Φ(x) withrespect to the center cD can be calculated by

1003817100381710038171003817Φ (x) minus cD1003817100381710038171003817 = sum

119894119895isinD

120572119894120572119895120581 (x119894 x119895)

minus 2sum

119894isinD

120572119894120581 (x119894 x) + 120581 (x x)

(10)

A modification of the original formulation of the one-class classification algorithm that consists of minimizing theapproximation error c

119899minus cD is [27 28]

120572 = arg min120572119894119894isinD

1003817100381710038171003817100381710038171003817100381710038171003817

1

119899

119899

sum

119894=1

Φ(x119894) minus sum

119894isinD

120572119894Φ(x119894)

1003817100381710038171003817100381710038171003817100381710038171003817

2

(11)

The final solution is given by

120572 = Kminus1120581 (12)

where K is the Grammatrix with (119894 119895)th entry 120581(x119894 x119895) and 120581

is the column vector with entries (1119899)sum119899119894=1

120581(x119896 x119894) 119896 isin D

In the online scheme at each time step there is a newsample Let 120572

119899denote the coefficients K

119899denote the Gram

matrix and 120581119899denote the vector at time step 119899 A criterion is

used to determine whether the new sample can be includedinto the dictionary A threshold 120583

0is preset for the datum

x119905at time step 119905 the coherence-based sparsification criterion

[29 30] is

120598119905= max119894isinD

1003816100381610038161003816120581 (x119894 x119905)1003816100381610038161003816 (13)

First Case (120598119905gt 1205830) In this case the new data Φ(x

119899+1) is not

included into the dictionaryThe GrammatrixK119899+1

= K119899 120581119899

changes online

120581119899+1

=1

119899 + 1(119899120581119899+ b)

120572119899+1

= Kminus1119899+1120581119899+1

=119899

119899 + 1120572119899+

1

119899 + 1Kminus1119899b

(14)

where b is the column vector with entries 120581(x119894 x119899+1

) 119894 isin D

Second Case (120598119905le 1205830) In this case the new data Φ(x

119899+1) is

included into the dictionaryD The Grammatrix K changes

K119899+1

= [

[

K119899

bb⊤ 120581 (x

119899+1 x119899+1

)]

]

(15)

By using Woodbury matrix identity

(119860 + 119880119862119881)minus1

= 119860minus1

minus 119860minus1

119880(119862minus1

+ 119881119860minus1

119880)minus1

119881119860minus1

(16)

Kminus1119899+1

can be calculated iteratively

Kminus1119899+1

= [

[

Kminus1119899

00⊤ 0

]

]

+1

120581 (x119899+1

x119899+1

) minus b⊤Kminus1119899b

times [

[

minusKminus1119899b

1]

]

[minusb⊤Kminus11198991]

(17)

The vector 120581119899+1

is updated from 120581119899

120581119899+1

=1

119899 + 1[119899120581119899+

120581119899+1

] (18)

with

120581119899+1

=

119899+1

sum

119894=1

120581 (x119899+1

x119894) (19)

Computing 120581119899+1

as (19) needs to save all the samples x119899+1119894=1

inmemory For conquering this issue it can compute as 120581

119899+1=

(119899 + 1)120581(x119899+1

x119899+1

) by considering an instant estimationTheupdate of 120572

119899+1from 120572

119899is

120572119899+1

=1

119899 + 1[119899120572119899+ Kminus1119899b

0]

minus1

(119899 + 1) (120581 (x119899+1

x119899+1

) minus b⊤Kminus1119899b)

times [

[

Kminus1119899b

1]

]

(119899b⊤120572119899+ b⊤Kminus1

119899b minus 120581119899+1

)

(20)

Based on (20) we have an online implementation of the one-class SVM learning phase

4 Abnormal Events Detection

In an abnormal event detection problem it is assumedthat a set of training frames 119868

1 119868

119899 (the positive class)

describing the normal behavior is obtained The generalarchitectures of abnormal detection are introduced below

The offline training strategy refers to the case where allthe training samples are learnt as one batch as shown inFigure 3(a) We propose two abnormal detection strategiesthe difference between these two strategies is the time whenthe dictionary is fixed These two strategies are shown inFigures 3(b) and 3(c) Strategy 1 is shown in Figure 3(b) Thetraining data are learnt one by oneWhen the training periodis finished the dictionary and the classifier are fixed Eachtest datum is classified based on the dictionary Figure 3(c)illustrates Strategy 2 The training procedure is the same asStrategy 1 But in the testing period the dictionary is updatedif the datum x

119894satisfies the dictionary update condition The

details of these two strategies are explained in the following

Journal of Electrical and Computer Engineering 5

Train offline

Test onlinem n minus m

(a) Strategy offline

Test online

Train online

m n minus m

Train

Dictionary fixed

(b) Strategy 1

Train

Dictionary fixed

Train and test online

Test onlinem n minus m

(c) Strategy 2

Figure 3 Offline and two online abnormal event detection strategies based on one-class SVM (a) Strategy offlineThe training data are learntas one batch offline (b) Strategy 1 The dictionary is fixed when all the training data are learnt (c) Strategy 2 The dictionary continues beingupdated through the testing period

Features selectionon original image

Learning step (online)Covariance

Covariance

optical flow

Classification

SVM train

(online training)

people walk

Detection step (online)One-class SVM

R

Origin

Abnormal eventdetection

people run

optical flow

x11 x12 middot middot middot

middot middot middot

middot middot middot

x1n

x21 x22 x2n

xn1 xn2 xnn

x11 x12 middot middot middot

middot middot middot

middot middot middot

x1n

x21 x22 x2n

xn1 xn2 xnn

⋱

⋱

Figure 4 Major processing states of the proposed abnormal frame event detection method The covariance of the frame is computed

6 Journal of Electrical and Computer Engineering

41 Strategy 1 In Strategy 1 the dictionary is updated merelythrough the training period

Step 1 The first step is calculating the covariance matrixdescriptor of training frames based on the image intensity andthe optical flow This step can be generalized as

(1198681OP1) (1198682OP2) (119868

119899OP119899) 997888rarr C

1C2 C

119899

(21)

where (I1OP1) (I2OP2) (119868

119899OP119899) are the image

intensity and the corresponding optical flow of the 1st to 119899thframe C

1C2 C

119899 are the covariance matrix descriptors

Step 2 The second step consists of applying one-class SVMon the small subset of extracted descriptor of the trainingnormal frames to obtain the support vectors Consider asubset C

119894119898

119894=1 1 le 119898 ≪ 119899 of data selected from the full

training sample set C119894119899

119894=1 without loss of generality and

assume that the first 119898 examples are chosen This set of 119898examples is called dictionary CD

C1C2 C

119898 1 le 119898 ≪ 119899

SVM997888rarr support vector 119878119901

1 1198781199012 119878119901

119900

(22)

where the set C1C2 C

119898 is the first119898 covariancematrix

descriptors of the training frames it is the original dictionaryCD In one-class SVM the majority of the training samplesdo not contribute to the definition of the decision functionThe entries of a monitory subset of the training samples1198781199011 1198781199012 119878119901

119900 119900 le 119898 are support vectors contributing

to the definition of the decision function

Step 3 After learning the dictionary CD which includesthe first 119898 1 le 119898 ≪ 119899 samples the training samplesC119898+1

C119898+2

C119899 are learned online via the technique

described in Section 3 This step can be generalized as

CDC119896 119898 lt 119896 le 119899SVM997888rarr support vector 119878119901

1 1198781199012 119878119901

119901

119900 le 119901 le 119899

CD = CD cup C119896 if 120598

119905ge 1205830

(23)

where CD is the dictionary obtained through Step 2 C119896is a

new sample in the remaining training dataset According tothe criterion introduced in Section 3 if the new sample C

119896

satisfies the dictionary updated condition it will be includedinto the dictionary CD

Step 4 Based on the dictionary and the classifier obtainedfrom the training frames the incoming frame sample C

119899+119897is

Table 2 AUC of abnormal event detection method of differentfeatures 119865 via original SVDD which learns training samples offlineStrategy 1 online one-class SVM and Strategy 2 online one-classSVMThe biggest value of each method is shown in bold

Features Area under ROCLawn Indoor Plaza

Training samples are learned offline1198651(6 times 6119889119906) 09426 08351 09323

1198652(6 times 6119889V) 09400 08358 09321

1198653(8 times 8) 09440 08375 09359

1198654(12 times 12) 09591 08440 09580

1198655(17 times 17) 09567 08649 09649

Strategy 11198651(6 times 6119889119906) 09399 08328 09343

1198652(6 times 6119889V) 09390 08355 09366

1198653(8 times 8) 09418 08377 09411

1198654(12 times 12) 09581 08457 09573

1198655(17 times 17) 09551 08628 09632

Strategy 21198651(6 times 6119889119906) 09427 08237 09288

1198652(6 times 6119889V) 09370 08241 09283

1198653(8 times 8) 09430 08274 09312

1198654(12 times 12) 09605 08331 09505

1198655(17 times 17) 09601 08495 09746

classifiedTheworkflow of Strategy 1 is shown in Figure 4 anddescribed by the following equation

119891 (C119899+119897) = sgn(1198772 minus

119899

sum

119894119895=1

120572119894120572119895120581 (C119894C119895)

+ 2sum

119894

120572119894120581 (C119894C119899+119897) minus 120581 (C

119899+119897C119899+119897))

= 1 119891 (C

119899+119897) ge 0

minus1 119891 (C119899+119897) lt 0

(24)

whereC119899+119897

is the covariance matrix descriptor of the (119899+ 119897)thframe needed to be classified and C

119894and C

119895are the samples

of the dictionary CD ldquo1rdquo corresponds to the normal frameldquominus1rdquo corresponds to the abnormal frame

42 Strategy 2 In this strategy the dictionary is updatedthrough both training and testing periodsThe feature extrac-tion step (Step 1) and the online training steps (Steps 2 and 3)are the same as the ones presented in Strategy 1 The testingstep is different The new coming datum which is detectedas normal but satisfies dictionary update condition shouldbe included into CD The dictionary is needed to be updatedthrough the testing period to include new samples

Step 4 Strategy 2 If the incoming frame sample C119899+119897

isclassified as normal (119891(C

119899+119897) = 1) the data is checked by

the criterion described in Section 3 When the data satisfies

Journal of Electrical and Computer Engineering 7

(a) Normal lawn scene (b) Abnormal lawn scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC lawn offline

094092

09088086084082

080 005 01 015 02

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

094092

09088086084082

080 005 01 015 02

ROC lawn Strategy 1

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 5 Detection results of the lawn scene (a) The detection result of one normal frame (b) The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the lawn scene results via one-class SVM All the training samples are learned together offlineThe biggest AUC value is 09591 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUC value is09581

the dictionary update criterion this testing sample will beincluded into the dictionary This step can be generalized bythe following equation

119891 (C119899+119897) = sgn(1198772 minus

119899

sum

119894119895=1

120572119894120572119895120581 (C119894C119895)

+ 2sum

119894

120572119894120581 (C119894C119899+119897) minus 120581 (C

119899+119897C119899+119897))

=

1 119891 (C119899+119897) ge 0

120598119905ge 1205830997888rarr CD = CD cup C

119899+119897

120598119905lt 1205830997888rarr CD = CD

minus1 119891 (C119899+119897) lt 0

(25)

5 Abnormal Detection Result

This section presents the results of experiments conducted toanalyze the performance of the proposedmethod A compet-itive performance through both Strategy 1 and Strategy 2 ofUMN [31] dataset is presented

51 Abnormal Visual Events Detection Strategy 1 The resultsof the proposed abnormal events detection method viaStrategy 1 online one-class SVM of UMN [31] dataset areshown below

The UMN dataset includes eleven video sequences ofthree different scenes (the lawn indoor and plaza) ofcrowded escape events The normal samples for training or

8 Journal of Electrical and Computer Engineering

(a) Normal indoor scene (b) Abnormal indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC indoor offline

085

08

075

07

065

0 01 02 03 04

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC indoor Strategy 1

085

08

075

07

065

0 01 02 03 04

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 6 Detection results of the indoor scene (a)The detection result of one normal frame (b)The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the indoor scene results via one-class SVM All the training samples are learned togetheroffline The biggest AUC value is 08649 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUCvalue is 08628

Table 3 The comparison of our proposed method with the state-of-the-art methods for the abnormal event detection in the UMNdataset

Method Area under ROCLawn Indoor Plaza

Social force [16] 096Optical flow [16] 084NN [17] 093SRC [17] 0995 0975 0964STCOG [18] 09362 07759 09661COV online (ours) 09605 08628 09746

for normal testing are the frames where the people are walk-ing in different directions The samples for abnormal testingare the frames where people are running The detectionresults of the lawn scene indoor scene and plaza scene areshown in Figures 5 6 and 7 respectively Gaussian kernelfor the Lie Group is used in these three scenes Differentvalues of120590 and penalty factor119862 are chosen the area under theROC curve is shown as a function of these parameters [32]The results show that taking covariance matrix as descriptorcan obtain satisfactory performance for abnormal detectionAnd also training the samples online can obtain similarlydetection performance as training all the samples offline

Journal of Electrical and Computer Engineering 9

(a) Normal plaza scene (b) Abnormal plaza scene

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

ROC plaza offline

False positive

True

pos

itive

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

False positive

True

pos

itive

ROC plaza Strategy 1

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 7 Detection results of the plaza scene (a) The detection result of one normal frame (b) The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the plaza scene results via one-class SVM All the training samples are learned together offlineThe biggest AUC value is 09649 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUC value is09632

Online one-class SVM is appropriate to detect abnormalvisual events There are 1431 frames in the lawn scene and480 normal frames are used for training In the offline strat-egy all the 480 frames covariancematrices should be saved inthememory In Strategy 1 100 frames covariancematrices areconsidered as the dictionary firstlyWhen119865

5-17times17 feature is

adopted to construct the covariance descriptor the varianceof Gaussian kernel is 120590 = 1 the preset threshold of thecriterion is 120583

0= 05 the dictionary size increases from 100

to 101 and the maximum accuracy of the detection resultsis 9169 In the indoor scene there are 2975 normal framesand 1057 abnormal frames In the plaza scene there are 1831normal frames and 286 abnormal framesTheprocesses of theexperiments are similar to the ones of the lawn scene Whenfeature vector is 119865

5-17 times 17 120590 = 1 120583

0= 05 the dictionary

size of these two scenes remain 100The online strategy keepsthe memory size almost unchanged when the size of trainingdataset increases

52 Abnormal Visual Events Detection Strategy 2 The resultsof the abnormal event detection method via Strategy 2 ofUMNdataset are shown as follows In the experiment processof the lawn scene 100 normal samples from the trainingsamples are learnt firstly and then the other 380 trainingdata are learnt online one by one After these two trainingsteps we can obtain the basic dictionary from the trainingsamples and also the classifier In the following testing stepthe dictionary is updated if the sample satisfies the dictionaryupdate criterion When a new sample is coming it is firstlydetected by the previous classifier If it is classified as anomaly

10 Journal of Electrical and Computer Engineering

094092

09088086084082

08

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC lawn Strategy 2

F1

F2

F3

F4

F5

(a) ROC Strategy 2 lawn scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

085

08

075

07

065

0 01 02 03 04

ROC indoor Strategy 2

F1

F2

F3

F4

F5

(b) ROC Strategy 2 indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC plaza Strategy 2

F1

F2

F3

F4

F5

098096094092

09088086084082

08

(c) ROC Strategy 2 plaza scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC train data learnt offline

LawnPlazaIndoor

(d) ROC train offline

Figure 8 ROC curve of UMN dataset (a) ROC curve of different features 119865 results via Strategy 2 of the lawn sceneThe biggest AUC value is09605 (b) Strategy 2 results of the indoor scene The biggest AUC value is 08495 (c) Strategy 2 results of the plaza scene The biggest AUCvalue is 09746 (d)The ROC curve of best performance of the lawn indoor and plaza scene when the training samples are learnt offlineThebiggest AUC values of the lawn indoor and plaza are 09591 08649 and 09649

the dictionary and the classifier are not changed Otherwiseif the sample is classified as a normal one the sparse criterionintroduced in Section 3 is used to check the correlationbetween the earlier dictionary and this new datum It willbe included into the dictionary when it satisfied the updatecondition The dictionary will be updated through the wholetesting period The other two scenes the indoor and plazascene are handled by the same methods When 119865

5-17 times 17

feature is adopted the variance of the Gaussian kernel is120590 = 1 and the preset threshold of the criterion is 120583

0= 05

and the dictionary size of the lawn indoor and plaza scene

is increased from 100 to 106 102 and 102 respectively TheROC curve of detection results of these three scenes is shownin Figures 8(a) 8(b) and 8(c) Besides the merit of savingmemory of Strategy 1 Strategy 2 also has the advantage ofadaptation to the long duration sequence

The results performances of offline strategy Strategy 1and Strategy 2 are shown in Table 2 The performances ofthese two strategies results are similar to that of the resultswhen all training samples are learnt together When 119865

4(12 times

12) or1198655(17times17) are chosen as the features to formcovariance

matrix descriptor the results have the best performance

Journal of Electrical and Computer Engineering 11

These two features are more abundant to include movementand intensity information

The result performances of the covariancematrix descrip-tor based online one-class SVM method and the state-of-the-art methods are shown in Table 3 The covariance matrixbased online abnormal frame detection method obtainscompetitive performance In general our method is betterthan others except sparse reconstruction cost (SRC) [17]in lawn scene and indoor scene In that paper multiscaleHOF is taken as a feature and a testing sample is classi-fied by its sparse reconstructor cost through a weightedlinear reconstruction of the overcomplete normal basis setBut computation of the HOF might takes more time thancalculating covariance By adopting the integral image [15]the covariance matrix descriptor of the subimage can becomputed conveniently So the covariance descriptor can beappropriately used to analyze the partial movement

6 Conclusions

A method for the abnormal event detection of the frameis proposed The method consists of covariance matrixdescriptor encoding the movement features and the onlinenonlinear one-class SVM classification method We havedeveloped two nonlinear one-class SVM based abnormalevent detection techniques that update the normal modelsof the surveillance video data in an online framework Theproposed algorithm has been tested on a video datasetyielding successful results to detect abnormal events

Acknowledgments

This work is partially supported by China ScholarshipCouncil of Chinese Government SURECAP CPER Projectfunded by Region Champagne-Ardenne and CAPSECCRCA FEDER platform

References

[1] F Jiang J Yuan S A Tsaftaris and A K Katsaggelos ldquoAnoma-lous video event detection using spatiotemporal contextrdquo Com-puter Vision and Image Understanding vol 115 no 3 pp 323ndash333 2011

[2] C Piciarelli C Micheloni and G L Foresti ldquoTrajectory-basedanomalous event detectionrdquo IEEE Transactions on Circuits andSystems for Video Technology vol 18 no 11 pp 1544ndash1554 2008

[3] C Piciarelli and G L Foresti ldquoOn-line trajectory clustering foranomalous events detectionrdquo Pattern Recognition Letters vol27 no 15 pp 1835ndash1842 2006

[4] T Xiang and S Gong ldquoIncremental and adaptive abnormalbehaviour detectionrdquo Computer Vision and Image Understand-ing vol 111 no 1 pp 59ndash73 2008

[5] S Gong and T Xiang ldquoRecognition of group activities usingdynamic probabilistic networksrdquo in Proceedings of the 9th IEEEInternational Conference on Computer Vision (ICCV rsquo03) pp742ndash749 October 2003

[6] J KimandKGrauman ldquoObserve locally infer globally a space-time MRF for detecting abnormal activities with incrementalupdatesrdquo in Proceedings of the IEEE Conference on Computer

Vision and Pattern Recognition (CVPR rsquo09) pp 2921ndash2928Miami Fla USA June 2009

[7] A Adam E Rivlin I Shimshoni and D Reinitz ldquoRobustreal-time unusual event detection using multiple fixed-locationmonitorsrdquo IEEE Transactions on Pattern Analysis and MachineIntelligence vol 30 no 3 pp 555ndash560 2008

[8] O Boiman andM Irani ldquoDetecting irregularities in images andin videordquo International Journal of Computer Vision vol 74 no1 pp 17ndash31 2007

[9] X Zhu Z Liu and J Zhang ldquoHuman activity clustering foronline anomaly detectionrdquo Journal of Computers vol 6 no 6pp 1071ndash1079 2011

[10] X Zhu and Z Liu ldquoHuman behavior clustering for anomalydetectionrdquo Frontiers of Computer Science in China vol 5 no3 pp 279ndash289 2011

[11] Y Benezeth P M Jodoin and V Saligrama ldquoAbnormalitydetection using low-level co-occurring eventsrdquo Pattern Recog-nition Letters vol 32 no 3 pp 423ndash431 2011

[12] Y Benezeth P M Jodoin V Saligrama and C RosenbergerldquoAbnormal events detection based on spatio-temporal co-occurencesrdquo in Proceedings of the IEEE Conference on ComputerVision and Pattern Recognition (CVPR rsquo09) pp 2458ndash2465Miami Fla USA June 2009

[13] A Mittal A Monnet and N Paragios ldquoScene modeling andchange detection in dynamic scenes a subspace approachrdquoComputer Vision and Image Understanding vol 113 no 1 pp63ndash79 2009

[14] B K P Horn and B G Schunck ldquoDetermining optical flowrdquoArtificial Intelligence vol 17 no 1ndash3 pp 185ndash203 1981

[15] O Tuzel F Porikli and P Meer ldquoRegion covariance afast descriptor for detection and classificationrdquo in ComputerVisionmdashECCV 2006 vol 3952 of Lecture Notes in ComputerScience pp 589ndash600 Springer New York NY USA 2006

[16] RMehran A Oyama andM Shah ldquoAbnormal crowd behaviordetection using social force modelrdquo in Proceedings of the IEEEConference on Computer Vision and Pattern Recognition (CVPRrsquo09) pp 935ndash942 Miami Fla USA June 2009

[17] Y Cong J Yuan and J Liu ldquoSparse reconstruction cost forabnormal event detectionrdquo in Proceedings of the IEEE Confer-ence on Computer Vision and Pattern Recognition (CVPR rsquo11)pp 3449ndash3456 Providence RI USA June 2011

[18] Y Shi Y Gao and R Wang ldquoReal-time abnormal eventdetection in complicated scenesrdquo in Proceedings of the 20thInternational Conference on Pattern Recognition (ICPR rsquo10) pp3653ndash3656 Istanbul Turkey August 2010

[19] B Scholkopf and A J Smola Learning with Kernels SupportVector Machines Regularization Optimization and BeyondMIT Press Boston Mass USA 2002

[20] B Hall Lie Groups Lie Algebras And Representations AnElementary Introduction vol 222 Springer New York NYUSA 2003

[21] C Gao F Li and C Shen ldquoResearch on lie group kernellearning algorithmrdquo Journal of Frontiers of Compter Science andTechnology vol 6 pp 1026ndash1038 2012

[22] V N Vapnik and A Lerner ldquoPattern recognition using general-ized portrait methodrdquo Automation and Remote Control vol 24pp 774ndash780 1963

[23] V Vapnik The Nature of Statistical Learning Theory SpringerNew York NY USA 2000

[24] D TaxOne-class classification [PhD thesis] Delft University ofTechnology 2001

12 Journal of Electrical and Computer Engineering

[25] D M Tax and R P Duin ldquoData domain description usingsupport vectorsrdquo in Proceedings of the European Symposiumon Artificial Neural Networks Computational Intelligence andMachine Learning (ESANN rsquo99) vol 99 pp 251ndash256 1999

[26] B Scholkopf J C Platt J Shawe-Taylor A J Smola and RC Williamson ldquoEstimating the support of a high-dimensionaldistributionrdquo Neural Computation vol 13 no 7 pp 1443ndash14712001

[27] Z Noumir P Honeine and C Richard ldquoOnline one-classmachines based on the coherence criterionrdquo in Proceedings ofthe 20th European Signal Processing Conference (EUSIPCO rsquo12)pp 664ndash668 2012

[28] Z Noumir P Honeine and C Richard ldquoOne-class machinesbased on the coherence criterionrdquo in Proceedings of the IEEEStatistical Signal Processing Workshop (SSP rsquo12) pp 600ndash6032012

[29] P Honeine ldquoOnline kernel principal component analysis areducedorder modelrdquo IEEE Transactions on Pattern Analysisand Machine Intelligence vol 34 no 9 pp 1814ndash1826 2012

[30] C Richard J C M Bermudez and P Honeine ldquoOnlineprediction of time series data with kernelsrdquo IEEE Transactionson Signal Processing vol 57 no 3 pp 1058ndash1067 2009

[31] UMN ldquoUnusual crowd activity dataset of university of Min-nesotardquo 2006 httpmhacsumneduproj eventsshtml

[32] J A Hanley and B J McNeil ldquoThe meaning and use of thearea under a receiver operating characteristic (ROC) curverdquoRadiology vol 143 no 1 pp 29ndash36 1982

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

4 Journal of Electrical and Computer Engineering

used in our work Let cD denotes a sparse model of the centerc119899= (1119899)sum

119899

119894=1Φ(x119894) by using a small subset of available

samples which is called dictionary

cD = sum

119894isinD

120572119894Φ(x119894) (9)

whereD sub 1 2 119899 and let119873D denote the cardinality ofthis subset xD The distance of a mapped datum Φ(x) withrespect to the center cD can be calculated by

1003817100381710038171003817Φ (x) minus cD1003817100381710038171003817 = sum

119894119895isinD

120572119894120572119895120581 (x119894 x119895)

minus 2sum

119894isinD

120572119894120581 (x119894 x) + 120581 (x x)

(10)

A modification of the original formulation of the one-class classification algorithm that consists of minimizing theapproximation error c

119899minus cD is [27 28]

120572 = arg min120572119894119894isinD

1003817100381710038171003817100381710038171003817100381710038171003817

1

119899

119899

sum

119894=1

Φ(x119894) minus sum

119894isinD

120572119894Φ(x119894)

1003817100381710038171003817100381710038171003817100381710038171003817

2

(11)

The final solution is given by

120572 = Kminus1120581 (12)

where K is the Grammatrix with (119894 119895)th entry 120581(x119894 x119895) and 120581

is the column vector with entries (1119899)sum119899119894=1

120581(x119896 x119894) 119896 isin D

In the online scheme at each time step there is a newsample Let 120572

119899denote the coefficients K

119899denote the Gram

matrix and 120581119899denote the vector at time step 119899 A criterion is

used to determine whether the new sample can be includedinto the dictionary A threshold 120583

0is preset for the datum

x119905at time step 119905 the coherence-based sparsification criterion

[29 30] is

120598119905= max119894isinD

1003816100381610038161003816120581 (x119894 x119905)1003816100381610038161003816 (13)

First Case (120598119905gt 1205830) In this case the new data Φ(x

119899+1) is not

included into the dictionaryThe GrammatrixK119899+1

= K119899 120581119899

changes online

120581119899+1

=1

119899 + 1(119899120581119899+ b)

120572119899+1

= Kminus1119899+1120581119899+1

=119899

119899 + 1120572119899+

1

119899 + 1Kminus1119899b

(14)

where b is the column vector with entries 120581(x119894 x119899+1

) 119894 isin D

Second Case (120598119905le 1205830) In this case the new data Φ(x

119899+1) is

included into the dictionaryD The Grammatrix K changes

K119899+1

= [

[

K119899

bb⊤ 120581 (x

119899+1 x119899+1

)]

]

(15)

By using Woodbury matrix identity

(119860 + 119880119862119881)minus1

= 119860minus1

minus 119860minus1

119880(119862minus1

+ 119881119860minus1

119880)minus1

119881119860minus1

(16)

Kminus1119899+1

can be calculated iteratively

Kminus1119899+1

= [

[

Kminus1119899

00⊤ 0

]

]

+1

120581 (x119899+1

x119899+1

) minus b⊤Kminus1119899b

times [

[

minusKminus1119899b

1]

]

[minusb⊤Kminus11198991]

(17)

The vector 120581119899+1

is updated from 120581119899

120581119899+1

=1

119899 + 1[119899120581119899+

120581119899+1

] (18)

with

120581119899+1

=

119899+1

sum

119894=1

120581 (x119899+1

x119894) (19)

Computing 120581119899+1

as (19) needs to save all the samples x119899+1119894=1

inmemory For conquering this issue it can compute as 120581

119899+1=

(119899 + 1)120581(x119899+1

x119899+1

) by considering an instant estimationTheupdate of 120572

119899+1from 120572

119899is

120572119899+1

=1

119899 + 1[119899120572119899+ Kminus1119899b

0]

minus1

(119899 + 1) (120581 (x119899+1

x119899+1

) minus b⊤Kminus1119899b)

times [

[

Kminus1119899b

1]

]

(119899b⊤120572119899+ b⊤Kminus1

119899b minus 120581119899+1

)

(20)

Based on (20) we have an online implementation of the one-class SVM learning phase

4 Abnormal Events Detection

In an abnormal event detection problem it is assumedthat a set of training frames 119868

1 119868

119899 (the positive class)

describing the normal behavior is obtained The generalarchitectures of abnormal detection are introduced below

The offline training strategy refers to the case where allthe training samples are learnt as one batch as shown inFigure 3(a) We propose two abnormal detection strategiesthe difference between these two strategies is the time whenthe dictionary is fixed These two strategies are shown inFigures 3(b) and 3(c) Strategy 1 is shown in Figure 3(b) Thetraining data are learnt one by oneWhen the training periodis finished the dictionary and the classifier are fixed Eachtest datum is classified based on the dictionary Figure 3(c)illustrates Strategy 2 The training procedure is the same asStrategy 1 But in the testing period the dictionary is updatedif the datum x

119894satisfies the dictionary update condition The

details of these two strategies are explained in the following

Journal of Electrical and Computer Engineering 5

Train offline

Test onlinem n minus m

(a) Strategy offline

Test online

Train online

m n minus m

Train

Dictionary fixed

(b) Strategy 1

Train

Dictionary fixed

Train and test online

Test onlinem n minus m

(c) Strategy 2

Figure 3 Offline and two online abnormal event detection strategies based on one-class SVM (a) Strategy offlineThe training data are learntas one batch offline (b) Strategy 1 The dictionary is fixed when all the training data are learnt (c) Strategy 2 The dictionary continues beingupdated through the testing period

Features selectionon original image

Learning step (online)Covariance

Covariance

optical flow

Classification

SVM train

(online training)

people walk

Detection step (online)One-class SVM

R

Origin

Abnormal eventdetection

people run

optical flow

x11 x12 middot middot middot

middot middot middot

middot middot middot

x1n

x21 x22 x2n

xn1 xn2 xnn

x11 x12 middot middot middot

middot middot middot

middot middot middot

x1n

x21 x22 x2n

xn1 xn2 xnn

⋱

⋱

Figure 4 Major processing states of the proposed abnormal frame event detection method The covariance of the frame is computed

6 Journal of Electrical and Computer Engineering

41 Strategy 1 In Strategy 1 the dictionary is updated merelythrough the training period

Step 1 The first step is calculating the covariance matrixdescriptor of training frames based on the image intensity andthe optical flow This step can be generalized as

(1198681OP1) (1198682OP2) (119868

119899OP119899) 997888rarr C

1C2 C

119899

(21)

where (I1OP1) (I2OP2) (119868

119899OP119899) are the image

intensity and the corresponding optical flow of the 1st to 119899thframe C

1C2 C

119899 are the covariance matrix descriptors

Step 2 The second step consists of applying one-class SVMon the small subset of extracted descriptor of the trainingnormal frames to obtain the support vectors Consider asubset C

119894119898

119894=1 1 le 119898 ≪ 119899 of data selected from the full

training sample set C119894119899

119894=1 without loss of generality and

assume that the first 119898 examples are chosen This set of 119898examples is called dictionary CD

C1C2 C

119898 1 le 119898 ≪ 119899

SVM997888rarr support vector 119878119901

1 1198781199012 119878119901

119900

(22)

where the set C1C2 C

119898 is the first119898 covariancematrix

descriptors of the training frames it is the original dictionaryCD In one-class SVM the majority of the training samplesdo not contribute to the definition of the decision functionThe entries of a monitory subset of the training samples1198781199011 1198781199012 119878119901

119900 119900 le 119898 are support vectors contributing

to the definition of the decision function

Step 3 After learning the dictionary CD which includesthe first 119898 1 le 119898 ≪ 119899 samples the training samplesC119898+1

C119898+2

C119899 are learned online via the technique

described in Section 3 This step can be generalized as

CDC119896 119898 lt 119896 le 119899SVM997888rarr support vector 119878119901

1 1198781199012 119878119901

119901

119900 le 119901 le 119899

CD = CD cup C119896 if 120598

119905ge 1205830

(23)

where CD is the dictionary obtained through Step 2 C119896is a

new sample in the remaining training dataset According tothe criterion introduced in Section 3 if the new sample C

119896

satisfies the dictionary updated condition it will be includedinto the dictionary CD

Step 4 Based on the dictionary and the classifier obtainedfrom the training frames the incoming frame sample C

119899+119897is

Table 2 AUC of abnormal event detection method of differentfeatures 119865 via original SVDD which learns training samples offlineStrategy 1 online one-class SVM and Strategy 2 online one-classSVMThe biggest value of each method is shown in bold

Features Area under ROCLawn Indoor Plaza

Training samples are learned offline1198651(6 times 6119889119906) 09426 08351 09323

1198652(6 times 6119889V) 09400 08358 09321

1198653(8 times 8) 09440 08375 09359

1198654(12 times 12) 09591 08440 09580

1198655(17 times 17) 09567 08649 09649

Strategy 11198651(6 times 6119889119906) 09399 08328 09343

1198652(6 times 6119889V) 09390 08355 09366

1198653(8 times 8) 09418 08377 09411

1198654(12 times 12) 09581 08457 09573

1198655(17 times 17) 09551 08628 09632

Strategy 21198651(6 times 6119889119906) 09427 08237 09288

1198652(6 times 6119889V) 09370 08241 09283

1198653(8 times 8) 09430 08274 09312

1198654(12 times 12) 09605 08331 09505

1198655(17 times 17) 09601 08495 09746

classifiedTheworkflow of Strategy 1 is shown in Figure 4 anddescribed by the following equation

119891 (C119899+119897) = sgn(1198772 minus

119899

sum

119894119895=1

120572119894120572119895120581 (C119894C119895)

+ 2sum

119894

120572119894120581 (C119894C119899+119897) minus 120581 (C

119899+119897C119899+119897))

= 1 119891 (C

119899+119897) ge 0

minus1 119891 (C119899+119897) lt 0

(24)

whereC119899+119897

is the covariance matrix descriptor of the (119899+ 119897)thframe needed to be classified and C

119894and C

119895are the samples

of the dictionary CD ldquo1rdquo corresponds to the normal frameldquominus1rdquo corresponds to the abnormal frame

42 Strategy 2 In this strategy the dictionary is updatedthrough both training and testing periodsThe feature extrac-tion step (Step 1) and the online training steps (Steps 2 and 3)are the same as the ones presented in Strategy 1 The testingstep is different The new coming datum which is detectedas normal but satisfies dictionary update condition shouldbe included into CD The dictionary is needed to be updatedthrough the testing period to include new samples

Step 4 Strategy 2 If the incoming frame sample C119899+119897

isclassified as normal (119891(C

119899+119897) = 1) the data is checked by

the criterion described in Section 3 When the data satisfies

Journal of Electrical and Computer Engineering 7

(a) Normal lawn scene (b) Abnormal lawn scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC lawn offline

094092

09088086084082

080 005 01 015 02

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

094092

09088086084082

080 005 01 015 02

ROC lawn Strategy 1

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 5 Detection results of the lawn scene (a) The detection result of one normal frame (b) The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the lawn scene results via one-class SVM All the training samples are learned together offlineThe biggest AUC value is 09591 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUC value is09581

the dictionary update criterion this testing sample will beincluded into the dictionary This step can be generalized bythe following equation

119891 (C119899+119897) = sgn(1198772 minus

119899

sum

119894119895=1

120572119894120572119895120581 (C119894C119895)

+ 2sum

119894

120572119894120581 (C119894C119899+119897) minus 120581 (C

119899+119897C119899+119897))

=

1 119891 (C119899+119897) ge 0

120598119905ge 1205830997888rarr CD = CD cup C

119899+119897

120598119905lt 1205830997888rarr CD = CD

minus1 119891 (C119899+119897) lt 0

(25)

5 Abnormal Detection Result

This section presents the results of experiments conducted toanalyze the performance of the proposedmethod A compet-itive performance through both Strategy 1 and Strategy 2 ofUMN [31] dataset is presented

51 Abnormal Visual Events Detection Strategy 1 The resultsof the proposed abnormal events detection method viaStrategy 1 online one-class SVM of UMN [31] dataset areshown below

The UMN dataset includes eleven video sequences ofthree different scenes (the lawn indoor and plaza) ofcrowded escape events The normal samples for training or

8 Journal of Electrical and Computer Engineering

(a) Normal indoor scene (b) Abnormal indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC indoor offline

085

08

075

07

065

0 01 02 03 04

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC indoor Strategy 1

085

08

075

07

065

0 01 02 03 04

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 6 Detection results of the indoor scene (a)The detection result of one normal frame (b)The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the indoor scene results via one-class SVM All the training samples are learned togetheroffline The biggest AUC value is 08649 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUCvalue is 08628

Table 3 The comparison of our proposed method with the state-of-the-art methods for the abnormal event detection in the UMNdataset

Method Area under ROCLawn Indoor Plaza

Social force [16] 096Optical flow [16] 084NN [17] 093SRC [17] 0995 0975 0964STCOG [18] 09362 07759 09661COV online (ours) 09605 08628 09746

for normal testing are the frames where the people are walk-ing in different directions The samples for abnormal testingare the frames where people are running The detectionresults of the lawn scene indoor scene and plaza scene areshown in Figures 5 6 and 7 respectively Gaussian kernelfor the Lie Group is used in these three scenes Differentvalues of120590 and penalty factor119862 are chosen the area under theROC curve is shown as a function of these parameters [32]The results show that taking covariance matrix as descriptorcan obtain satisfactory performance for abnormal detectionAnd also training the samples online can obtain similarlydetection performance as training all the samples offline

Journal of Electrical and Computer Engineering 9

(a) Normal plaza scene (b) Abnormal plaza scene

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

ROC plaza offline

False positive

True

pos

itive

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

False positive

True

pos

itive

ROC plaza Strategy 1

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 7 Detection results of the plaza scene (a) The detection result of one normal frame (b) The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the plaza scene results via one-class SVM All the training samples are learned together offlineThe biggest AUC value is 09649 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUC value is09632

Online one-class SVM is appropriate to detect abnormalvisual events There are 1431 frames in the lawn scene and480 normal frames are used for training In the offline strat-egy all the 480 frames covariancematrices should be saved inthememory In Strategy 1 100 frames covariancematrices areconsidered as the dictionary firstlyWhen119865

5-17times17 feature is

adopted to construct the covariance descriptor the varianceof Gaussian kernel is 120590 = 1 the preset threshold of thecriterion is 120583

0= 05 the dictionary size increases from 100

to 101 and the maximum accuracy of the detection resultsis 9169 In the indoor scene there are 2975 normal framesand 1057 abnormal frames In the plaza scene there are 1831normal frames and 286 abnormal framesTheprocesses of theexperiments are similar to the ones of the lawn scene Whenfeature vector is 119865

5-17 times 17 120590 = 1 120583

0= 05 the dictionary

size of these two scenes remain 100The online strategy keepsthe memory size almost unchanged when the size of trainingdataset increases

52 Abnormal Visual Events Detection Strategy 2 The resultsof the abnormal event detection method via Strategy 2 ofUMNdataset are shown as follows In the experiment processof the lawn scene 100 normal samples from the trainingsamples are learnt firstly and then the other 380 trainingdata are learnt online one by one After these two trainingsteps we can obtain the basic dictionary from the trainingsamples and also the classifier In the following testing stepthe dictionary is updated if the sample satisfies the dictionaryupdate criterion When a new sample is coming it is firstlydetected by the previous classifier If it is classified as anomaly

10 Journal of Electrical and Computer Engineering

094092

09088086084082

08

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC lawn Strategy 2

F1

F2

F3

F4

F5

(a) ROC Strategy 2 lawn scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

085

08

075

07

065

0 01 02 03 04

ROC indoor Strategy 2

F1

F2

F3

F4

F5

(b) ROC Strategy 2 indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC plaza Strategy 2

F1

F2

F3

F4

F5

098096094092

09088086084082

08

(c) ROC Strategy 2 plaza scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC train data learnt offline

LawnPlazaIndoor

(d) ROC train offline

Figure 8 ROC curve of UMN dataset (a) ROC curve of different features 119865 results via Strategy 2 of the lawn sceneThe biggest AUC value is09605 (b) Strategy 2 results of the indoor scene The biggest AUC value is 08495 (c) Strategy 2 results of the plaza scene The biggest AUCvalue is 09746 (d)The ROC curve of best performance of the lawn indoor and plaza scene when the training samples are learnt offlineThebiggest AUC values of the lawn indoor and plaza are 09591 08649 and 09649

the dictionary and the classifier are not changed Otherwiseif the sample is classified as a normal one the sparse criterionintroduced in Section 3 is used to check the correlationbetween the earlier dictionary and this new datum It willbe included into the dictionary when it satisfied the updatecondition The dictionary will be updated through the wholetesting period The other two scenes the indoor and plazascene are handled by the same methods When 119865

5-17 times 17

feature is adopted the variance of the Gaussian kernel is120590 = 1 and the preset threshold of the criterion is 120583

0= 05

and the dictionary size of the lawn indoor and plaza scene

is increased from 100 to 106 102 and 102 respectively TheROC curve of detection results of these three scenes is shownin Figures 8(a) 8(b) and 8(c) Besides the merit of savingmemory of Strategy 1 Strategy 2 also has the advantage ofadaptation to the long duration sequence

The results performances of offline strategy Strategy 1and Strategy 2 are shown in Table 2 The performances ofthese two strategies results are similar to that of the resultswhen all training samples are learnt together When 119865

4(12 times

12) or1198655(17times17) are chosen as the features to formcovariance

matrix descriptor the results have the best performance

Journal of Electrical and Computer Engineering 11

These two features are more abundant to include movementand intensity information

The result performances of the covariancematrix descrip-tor based online one-class SVM method and the state-of-the-art methods are shown in Table 3 The covariance matrixbased online abnormal frame detection method obtainscompetitive performance In general our method is betterthan others except sparse reconstruction cost (SRC) [17]in lawn scene and indoor scene In that paper multiscaleHOF is taken as a feature and a testing sample is classi-fied by its sparse reconstructor cost through a weightedlinear reconstruction of the overcomplete normal basis setBut computation of the HOF might takes more time thancalculating covariance By adopting the integral image [15]the covariance matrix descriptor of the subimage can becomputed conveniently So the covariance descriptor can beappropriately used to analyze the partial movement

6 Conclusions

A method for the abnormal event detection of the frameis proposed The method consists of covariance matrixdescriptor encoding the movement features and the onlinenonlinear one-class SVM classification method We havedeveloped two nonlinear one-class SVM based abnormalevent detection techniques that update the normal modelsof the surveillance video data in an online framework Theproposed algorithm has been tested on a video datasetyielding successful results to detect abnormal events

Acknowledgments

This work is partially supported by China ScholarshipCouncil of Chinese Government SURECAP CPER Projectfunded by Region Champagne-Ardenne and CAPSECCRCA FEDER platform

References

[1] F Jiang J Yuan S A Tsaftaris and A K Katsaggelos ldquoAnoma-lous video event detection using spatiotemporal contextrdquo Com-puter Vision and Image Understanding vol 115 no 3 pp 323ndash333 2011

[2] C Piciarelli C Micheloni and G L Foresti ldquoTrajectory-basedanomalous event detectionrdquo IEEE Transactions on Circuits andSystems for Video Technology vol 18 no 11 pp 1544ndash1554 2008

[3] C Piciarelli and G L Foresti ldquoOn-line trajectory clustering foranomalous events detectionrdquo Pattern Recognition Letters vol27 no 15 pp 1835ndash1842 2006

[4] T Xiang and S Gong ldquoIncremental and adaptive abnormalbehaviour detectionrdquo Computer Vision and Image Understand-ing vol 111 no 1 pp 59ndash73 2008

[5] S Gong and T Xiang ldquoRecognition of group activities usingdynamic probabilistic networksrdquo in Proceedings of the 9th IEEEInternational Conference on Computer Vision (ICCV rsquo03) pp742ndash749 October 2003

[6] J KimandKGrauman ldquoObserve locally infer globally a space-time MRF for detecting abnormal activities with incrementalupdatesrdquo in Proceedings of the IEEE Conference on Computer

Vision and Pattern Recognition (CVPR rsquo09) pp 2921ndash2928Miami Fla USA June 2009

[7] A Adam E Rivlin I Shimshoni and D Reinitz ldquoRobustreal-time unusual event detection using multiple fixed-locationmonitorsrdquo IEEE Transactions on Pattern Analysis and MachineIntelligence vol 30 no 3 pp 555ndash560 2008

[8] O Boiman andM Irani ldquoDetecting irregularities in images andin videordquo International Journal of Computer Vision vol 74 no1 pp 17ndash31 2007

[9] X Zhu Z Liu and J Zhang ldquoHuman activity clustering foronline anomaly detectionrdquo Journal of Computers vol 6 no 6pp 1071ndash1079 2011

[10] X Zhu and Z Liu ldquoHuman behavior clustering for anomalydetectionrdquo Frontiers of Computer Science in China vol 5 no3 pp 279ndash289 2011

[11] Y Benezeth P M Jodoin and V Saligrama ldquoAbnormalitydetection using low-level co-occurring eventsrdquo Pattern Recog-nition Letters vol 32 no 3 pp 423ndash431 2011

[12] Y Benezeth P M Jodoin V Saligrama and C RosenbergerldquoAbnormal events detection based on spatio-temporal co-occurencesrdquo in Proceedings of the IEEE Conference on ComputerVision and Pattern Recognition (CVPR rsquo09) pp 2458ndash2465Miami Fla USA June 2009

[13] A Mittal A Monnet and N Paragios ldquoScene modeling andchange detection in dynamic scenes a subspace approachrdquoComputer Vision and Image Understanding vol 113 no 1 pp63ndash79 2009

[14] B K P Horn and B G Schunck ldquoDetermining optical flowrdquoArtificial Intelligence vol 17 no 1ndash3 pp 185ndash203 1981

[15] O Tuzel F Porikli and P Meer ldquoRegion covariance afast descriptor for detection and classificationrdquo in ComputerVisionmdashECCV 2006 vol 3952 of Lecture Notes in ComputerScience pp 589ndash600 Springer New York NY USA 2006

[16] RMehran A Oyama andM Shah ldquoAbnormal crowd behaviordetection using social force modelrdquo in Proceedings of the IEEEConference on Computer Vision and Pattern Recognition (CVPRrsquo09) pp 935ndash942 Miami Fla USA June 2009

[17] Y Cong J Yuan and J Liu ldquoSparse reconstruction cost forabnormal event detectionrdquo in Proceedings of the IEEE Confer-ence on Computer Vision and Pattern Recognition (CVPR rsquo11)pp 3449ndash3456 Providence RI USA June 2011

[18] Y Shi Y Gao and R Wang ldquoReal-time abnormal eventdetection in complicated scenesrdquo in Proceedings of the 20thInternational Conference on Pattern Recognition (ICPR rsquo10) pp3653ndash3656 Istanbul Turkey August 2010

[19] B Scholkopf and A J Smola Learning with Kernels SupportVector Machines Regularization Optimization and BeyondMIT Press Boston Mass USA 2002

[20] B Hall Lie Groups Lie Algebras And Representations AnElementary Introduction vol 222 Springer New York NYUSA 2003

[21] C Gao F Li and C Shen ldquoResearch on lie group kernellearning algorithmrdquo Journal of Frontiers of Compter Science andTechnology vol 6 pp 1026ndash1038 2012

[22] V N Vapnik and A Lerner ldquoPattern recognition using general-ized portrait methodrdquo Automation and Remote Control vol 24pp 774ndash780 1963

[23] V Vapnik The Nature of Statistical Learning Theory SpringerNew York NY USA 2000

[24] D TaxOne-class classification [PhD thesis] Delft University ofTechnology 2001

12 Journal of Electrical and Computer Engineering

[25] D M Tax and R P Duin ldquoData domain description usingsupport vectorsrdquo in Proceedings of the European Symposiumon Artificial Neural Networks Computational Intelligence andMachine Learning (ESANN rsquo99) vol 99 pp 251ndash256 1999

[26] B Scholkopf J C Platt J Shawe-Taylor A J Smola and RC Williamson ldquoEstimating the support of a high-dimensionaldistributionrdquo Neural Computation vol 13 no 7 pp 1443ndash14712001

[27] Z Noumir P Honeine and C Richard ldquoOnline one-classmachines based on the coherence criterionrdquo in Proceedings ofthe 20th European Signal Processing Conference (EUSIPCO rsquo12)pp 664ndash668 2012

[28] Z Noumir P Honeine and C Richard ldquoOne-class machinesbased on the coherence criterionrdquo in Proceedings of the IEEEStatistical Signal Processing Workshop (SSP rsquo12) pp 600ndash6032012

[29] P Honeine ldquoOnline kernel principal component analysis areducedorder modelrdquo IEEE Transactions on Pattern Analysisand Machine Intelligence vol 34 no 9 pp 1814ndash1826 2012

[30] C Richard J C M Bermudez and P Honeine ldquoOnlineprediction of time series data with kernelsrdquo IEEE Transactionson Signal Processing vol 57 no 3 pp 1058ndash1067 2009

[31] UMN ldquoUnusual crowd activity dataset of university of Min-nesotardquo 2006 httpmhacsumneduproj eventsshtml

[32] J A Hanley and B J McNeil ldquoThe meaning and use of thearea under a receiver operating characteristic (ROC) curverdquoRadiology vol 143 no 1 pp 29ndash36 1982

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Journal of Electrical and Computer Engineering 5

Train offline

Test onlinem n minus m

(a) Strategy offline

Test online

Train online

m n minus m

Train

Dictionary fixed

(b) Strategy 1

Train

Dictionary fixed

Train and test online

Test onlinem n minus m

(c) Strategy 2

Figure 3 Offline and two online abnormal event detection strategies based on one-class SVM (a) Strategy offlineThe training data are learntas one batch offline (b) Strategy 1 The dictionary is fixed when all the training data are learnt (c) Strategy 2 The dictionary continues beingupdated through the testing period

Features selectionon original image

Learning step (online)Covariance

Covariance

optical flow

Classification

SVM train

(online training)

people walk

Detection step (online)One-class SVM

R

Origin

Abnormal eventdetection

people run

optical flow

x11 x12 middot middot middot

middot middot middot

middot middot middot

x1n

x21 x22 x2n

xn1 xn2 xnn

x11 x12 middot middot middot

middot middot middot

middot middot middot

x1n

x21 x22 x2n

xn1 xn2 xnn

⋱

⋱

Figure 4 Major processing states of the proposed abnormal frame event detection method The covariance of the frame is computed

6 Journal of Electrical and Computer Engineering

41 Strategy 1 In Strategy 1 the dictionary is updated merelythrough the training period

Step 1 The first step is calculating the covariance matrixdescriptor of training frames based on the image intensity andthe optical flow This step can be generalized as

(1198681OP1) (1198682OP2) (119868

119899OP119899) 997888rarr C

1C2 C

119899

(21)

where (I1OP1) (I2OP2) (119868

119899OP119899) are the image

intensity and the corresponding optical flow of the 1st to 119899thframe C

1C2 C

119899 are the covariance matrix descriptors

Step 2 The second step consists of applying one-class SVMon the small subset of extracted descriptor of the trainingnormal frames to obtain the support vectors Consider asubset C

119894119898

119894=1 1 le 119898 ≪ 119899 of data selected from the full

training sample set C119894119899

119894=1 without loss of generality and

assume that the first 119898 examples are chosen This set of 119898examples is called dictionary CD

C1C2 C

119898 1 le 119898 ≪ 119899

SVM997888rarr support vector 119878119901

1 1198781199012 119878119901

119900

(22)

where the set C1C2 C

119898 is the first119898 covariancematrix

descriptors of the training frames it is the original dictionaryCD In one-class SVM the majority of the training samplesdo not contribute to the definition of the decision functionThe entries of a monitory subset of the training samples1198781199011 1198781199012 119878119901

119900 119900 le 119898 are support vectors contributing

to the definition of the decision function

Step 3 After learning the dictionary CD which includesthe first 119898 1 le 119898 ≪ 119899 samples the training samplesC119898+1

C119898+2

C119899 are learned online via the technique

described in Section 3 This step can be generalized as

CDC119896 119898 lt 119896 le 119899SVM997888rarr support vector 119878119901

1 1198781199012 119878119901

119901

119900 le 119901 le 119899

CD = CD cup C119896 if 120598

119905ge 1205830

(23)

where CD is the dictionary obtained through Step 2 C119896is a

new sample in the remaining training dataset According tothe criterion introduced in Section 3 if the new sample C

119896

satisfies the dictionary updated condition it will be includedinto the dictionary CD

Step 4 Based on the dictionary and the classifier obtainedfrom the training frames the incoming frame sample C

119899+119897is

Table 2 AUC of abnormal event detection method of differentfeatures 119865 via original SVDD which learns training samples offlineStrategy 1 online one-class SVM and Strategy 2 online one-classSVMThe biggest value of each method is shown in bold

Features Area under ROCLawn Indoor Plaza

Training samples are learned offline1198651(6 times 6119889119906) 09426 08351 09323

1198652(6 times 6119889V) 09400 08358 09321

1198653(8 times 8) 09440 08375 09359

1198654(12 times 12) 09591 08440 09580

1198655(17 times 17) 09567 08649 09649

Strategy 11198651(6 times 6119889119906) 09399 08328 09343

1198652(6 times 6119889V) 09390 08355 09366

1198653(8 times 8) 09418 08377 09411

1198654(12 times 12) 09581 08457 09573

1198655(17 times 17) 09551 08628 09632

Strategy 21198651(6 times 6119889119906) 09427 08237 09288

1198652(6 times 6119889V) 09370 08241 09283

1198653(8 times 8) 09430 08274 09312

1198654(12 times 12) 09605 08331 09505

1198655(17 times 17) 09601 08495 09746

classifiedTheworkflow of Strategy 1 is shown in Figure 4 anddescribed by the following equation

119891 (C119899+119897) = sgn(1198772 minus

119899

sum

119894119895=1

120572119894120572119895120581 (C119894C119895)

+ 2sum

119894

120572119894120581 (C119894C119899+119897) minus 120581 (C

119899+119897C119899+119897))

= 1 119891 (C

119899+119897) ge 0

minus1 119891 (C119899+119897) lt 0

(24)

whereC119899+119897

is the covariance matrix descriptor of the (119899+ 119897)thframe needed to be classified and C

119894and C

119895are the samples

of the dictionary CD ldquo1rdquo corresponds to the normal frameldquominus1rdquo corresponds to the abnormal frame

42 Strategy 2 In this strategy the dictionary is updatedthrough both training and testing periodsThe feature extrac-tion step (Step 1) and the online training steps (Steps 2 and 3)are the same as the ones presented in Strategy 1 The testingstep is different The new coming datum which is detectedas normal but satisfies dictionary update condition shouldbe included into CD The dictionary is needed to be updatedthrough the testing period to include new samples

Step 4 Strategy 2 If the incoming frame sample C119899+119897

isclassified as normal (119891(C

119899+119897) = 1) the data is checked by

the criterion described in Section 3 When the data satisfies

Journal of Electrical and Computer Engineering 7

(a) Normal lawn scene (b) Abnormal lawn scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC lawn offline

094092

09088086084082

080 005 01 015 02

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

094092

09088086084082

080 005 01 015 02

ROC lawn Strategy 1

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 5 Detection results of the lawn scene (a) The detection result of one normal frame (b) The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the lawn scene results via one-class SVM All the training samples are learned together offlineThe biggest AUC value is 09591 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUC value is09581

the dictionary update criterion this testing sample will beincluded into the dictionary This step can be generalized bythe following equation

119891 (C119899+119897) = sgn(1198772 minus

119899

sum

119894119895=1

120572119894120572119895120581 (C119894C119895)

+ 2sum

119894

120572119894120581 (C119894C119899+119897) minus 120581 (C

119899+119897C119899+119897))

=

1 119891 (C119899+119897) ge 0

120598119905ge 1205830997888rarr CD = CD cup C

119899+119897

120598119905lt 1205830997888rarr CD = CD

minus1 119891 (C119899+119897) lt 0

(25)

5 Abnormal Detection Result

This section presents the results of experiments conducted toanalyze the performance of the proposedmethod A compet-itive performance through both Strategy 1 and Strategy 2 ofUMN [31] dataset is presented

51 Abnormal Visual Events Detection Strategy 1 The resultsof the proposed abnormal events detection method viaStrategy 1 online one-class SVM of UMN [31] dataset areshown below

The UMN dataset includes eleven video sequences ofthree different scenes (the lawn indoor and plaza) ofcrowded escape events The normal samples for training or

8 Journal of Electrical and Computer Engineering

(a) Normal indoor scene (b) Abnormal indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC indoor offline

085

08

075

07

065

0 01 02 03 04

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC indoor Strategy 1

085

08

075

07

065

0 01 02 03 04

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 6 Detection results of the indoor scene (a)The detection result of one normal frame (b)The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the indoor scene results via one-class SVM All the training samples are learned togetheroffline The biggest AUC value is 08649 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUCvalue is 08628

Table 3 The comparison of our proposed method with the state-of-the-art methods for the abnormal event detection in the UMNdataset

Method Area under ROCLawn Indoor Plaza

Social force [16] 096Optical flow [16] 084NN [17] 093SRC [17] 0995 0975 0964STCOG [18] 09362 07759 09661COV online (ours) 09605 08628 09746

for normal testing are the frames where the people are walk-ing in different directions The samples for abnormal testingare the frames where people are running The detectionresults of the lawn scene indoor scene and plaza scene areshown in Figures 5 6 and 7 respectively Gaussian kernelfor the Lie Group is used in these three scenes Differentvalues of120590 and penalty factor119862 are chosen the area under theROC curve is shown as a function of these parameters [32]The results show that taking covariance matrix as descriptorcan obtain satisfactory performance for abnormal detectionAnd also training the samples online can obtain similarlydetection performance as training all the samples offline

Journal of Electrical and Computer Engineering 9

(a) Normal plaza scene (b) Abnormal plaza scene

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

ROC plaza offline

False positive

True

pos

itive

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

False positive

True

pos

itive

ROC plaza Strategy 1

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 7 Detection results of the plaza scene (a) The detection result of one normal frame (b) The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the plaza scene results via one-class SVM All the training samples are learned together offlineThe biggest AUC value is 09649 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUC value is09632

Online one-class SVM is appropriate to detect abnormalvisual events There are 1431 frames in the lawn scene and480 normal frames are used for training In the offline strat-egy all the 480 frames covariancematrices should be saved inthememory In Strategy 1 100 frames covariancematrices areconsidered as the dictionary firstlyWhen119865

5-17times17 feature is

adopted to construct the covariance descriptor the varianceof Gaussian kernel is 120590 = 1 the preset threshold of thecriterion is 120583

0= 05 the dictionary size increases from 100

to 101 and the maximum accuracy of the detection resultsis 9169 In the indoor scene there are 2975 normal framesand 1057 abnormal frames In the plaza scene there are 1831normal frames and 286 abnormal framesTheprocesses of theexperiments are similar to the ones of the lawn scene Whenfeature vector is 119865

5-17 times 17 120590 = 1 120583

0= 05 the dictionary

size of these two scenes remain 100The online strategy keepsthe memory size almost unchanged when the size of trainingdataset increases

52 Abnormal Visual Events Detection Strategy 2 The resultsof the abnormal event detection method via Strategy 2 ofUMNdataset are shown as follows In the experiment processof the lawn scene 100 normal samples from the trainingsamples are learnt firstly and then the other 380 trainingdata are learnt online one by one After these two trainingsteps we can obtain the basic dictionary from the trainingsamples and also the classifier In the following testing stepthe dictionary is updated if the sample satisfies the dictionaryupdate criterion When a new sample is coming it is firstlydetected by the previous classifier If it is classified as anomaly

10 Journal of Electrical and Computer Engineering

094092

09088086084082

08

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC lawn Strategy 2

F1

F2

F3

F4

F5

(a) ROC Strategy 2 lawn scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

085

08

075

07

065

0 01 02 03 04

ROC indoor Strategy 2

F1

F2

F3

F4

F5

(b) ROC Strategy 2 indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC plaza Strategy 2

F1

F2

F3

F4

F5

098096094092

09088086084082

08

(c) ROC Strategy 2 plaza scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC train data learnt offline

LawnPlazaIndoor

(d) ROC train offline

Figure 8 ROC curve of UMN dataset (a) ROC curve of different features 119865 results via Strategy 2 of the lawn sceneThe biggest AUC value is09605 (b) Strategy 2 results of the indoor scene The biggest AUC value is 08495 (c) Strategy 2 results of the plaza scene The biggest AUCvalue is 09746 (d)The ROC curve of best performance of the lawn indoor and plaza scene when the training samples are learnt offlineThebiggest AUC values of the lawn indoor and plaza are 09591 08649 and 09649

the dictionary and the classifier are not changed Otherwiseif the sample is classified as a normal one the sparse criterionintroduced in Section 3 is used to check the correlationbetween the earlier dictionary and this new datum It willbe included into the dictionary when it satisfied the updatecondition The dictionary will be updated through the wholetesting period The other two scenes the indoor and plazascene are handled by the same methods When 119865

5-17 times 17

feature is adopted the variance of the Gaussian kernel is120590 = 1 and the preset threshold of the criterion is 120583

0= 05

and the dictionary size of the lawn indoor and plaza scene

is increased from 100 to 106 102 and 102 respectively TheROC curve of detection results of these three scenes is shownin Figures 8(a) 8(b) and 8(c) Besides the merit of savingmemory of Strategy 1 Strategy 2 also has the advantage ofadaptation to the long duration sequence

The results performances of offline strategy Strategy 1and Strategy 2 are shown in Table 2 The performances ofthese two strategies results are similar to that of the resultswhen all training samples are learnt together When 119865

4(12 times

12) or1198655(17times17) are chosen as the features to formcovariance

matrix descriptor the results have the best performance

Journal of Electrical and Computer Engineering 11

These two features are more abundant to include movementand intensity information

The result performances of the covariancematrix descrip-tor based online one-class SVM method and the state-of-the-art methods are shown in Table 3 The covariance matrixbased online abnormal frame detection method obtainscompetitive performance In general our method is betterthan others except sparse reconstruction cost (SRC) [17]in lawn scene and indoor scene In that paper multiscaleHOF is taken as a feature and a testing sample is classi-fied by its sparse reconstructor cost through a weightedlinear reconstruction of the overcomplete normal basis setBut computation of the HOF might takes more time thancalculating covariance By adopting the integral image [15]the covariance matrix descriptor of the subimage can becomputed conveniently So the covariance descriptor can beappropriately used to analyze the partial movement

6 Conclusions

A method for the abnormal event detection of the frameis proposed The method consists of covariance matrixdescriptor encoding the movement features and the onlinenonlinear one-class SVM classification method We havedeveloped two nonlinear one-class SVM based abnormalevent detection techniques that update the normal modelsof the surveillance video data in an online framework Theproposed algorithm has been tested on a video datasetyielding successful results to detect abnormal events

Acknowledgments

This work is partially supported by China ScholarshipCouncil of Chinese Government SURECAP CPER Projectfunded by Region Champagne-Ardenne and CAPSECCRCA FEDER platform

References

[1] F Jiang J Yuan S A Tsaftaris and A K Katsaggelos ldquoAnoma-lous video event detection using spatiotemporal contextrdquo Com-puter Vision and Image Understanding vol 115 no 3 pp 323ndash333 2011

[2] C Piciarelli C Micheloni and G L Foresti ldquoTrajectory-basedanomalous event detectionrdquo IEEE Transactions on Circuits andSystems for Video Technology vol 18 no 11 pp 1544ndash1554 2008

[3] C Piciarelli and G L Foresti ldquoOn-line trajectory clustering foranomalous events detectionrdquo Pattern Recognition Letters vol27 no 15 pp 1835ndash1842 2006

[4] T Xiang and S Gong ldquoIncremental and adaptive abnormalbehaviour detectionrdquo Computer Vision and Image Understand-ing vol 111 no 1 pp 59ndash73 2008

[5] S Gong and T Xiang ldquoRecognition of group activities usingdynamic probabilistic networksrdquo in Proceedings of the 9th IEEEInternational Conference on Computer Vision (ICCV rsquo03) pp742ndash749 October 2003

[6] J KimandKGrauman ldquoObserve locally infer globally a space-time MRF for detecting abnormal activities with incrementalupdatesrdquo in Proceedings of the IEEE Conference on Computer

Vision and Pattern Recognition (CVPR rsquo09) pp 2921ndash2928Miami Fla USA June 2009

[7] A Adam E Rivlin I Shimshoni and D Reinitz ldquoRobustreal-time unusual event detection using multiple fixed-locationmonitorsrdquo IEEE Transactions on Pattern Analysis and MachineIntelligence vol 30 no 3 pp 555ndash560 2008

[8] O Boiman andM Irani ldquoDetecting irregularities in images andin videordquo International Journal of Computer Vision vol 74 no1 pp 17ndash31 2007

[9] X Zhu Z Liu and J Zhang ldquoHuman activity clustering foronline anomaly detectionrdquo Journal of Computers vol 6 no 6pp 1071ndash1079 2011

[10] X Zhu and Z Liu ldquoHuman behavior clustering for anomalydetectionrdquo Frontiers of Computer Science in China vol 5 no3 pp 279ndash289 2011

[11] Y Benezeth P M Jodoin and V Saligrama ldquoAbnormalitydetection using low-level co-occurring eventsrdquo Pattern Recog-nition Letters vol 32 no 3 pp 423ndash431 2011

[12] Y Benezeth P M Jodoin V Saligrama and C RosenbergerldquoAbnormal events detection based on spatio-temporal co-occurencesrdquo in Proceedings of the IEEE Conference on ComputerVision and Pattern Recognition (CVPR rsquo09) pp 2458ndash2465Miami Fla USA June 2009

[13] A Mittal A Monnet and N Paragios ldquoScene modeling andchange detection in dynamic scenes a subspace approachrdquoComputer Vision and Image Understanding vol 113 no 1 pp63ndash79 2009

[14] B K P Horn and B G Schunck ldquoDetermining optical flowrdquoArtificial Intelligence vol 17 no 1ndash3 pp 185ndash203 1981

[15] O Tuzel F Porikli and P Meer ldquoRegion covariance afast descriptor for detection and classificationrdquo in ComputerVisionmdashECCV 2006 vol 3952 of Lecture Notes in ComputerScience pp 589ndash600 Springer New York NY USA 2006

[16] RMehran A Oyama andM Shah ldquoAbnormal crowd behaviordetection using social force modelrdquo in Proceedings of the IEEEConference on Computer Vision and Pattern Recognition (CVPRrsquo09) pp 935ndash942 Miami Fla USA June 2009

[17] Y Cong J Yuan and J Liu ldquoSparse reconstruction cost forabnormal event detectionrdquo in Proceedings of the IEEE Confer-ence on Computer Vision and Pattern Recognition (CVPR rsquo11)pp 3449ndash3456 Providence RI USA June 2011

[18] Y Shi Y Gao and R Wang ldquoReal-time abnormal eventdetection in complicated scenesrdquo in Proceedings of the 20thInternational Conference on Pattern Recognition (ICPR rsquo10) pp3653ndash3656 Istanbul Turkey August 2010

[19] B Scholkopf and A J Smola Learning with Kernels SupportVector Machines Regularization Optimization and BeyondMIT Press Boston Mass USA 2002

[20] B Hall Lie Groups Lie Algebras And Representations AnElementary Introduction vol 222 Springer New York NYUSA 2003

[21] C Gao F Li and C Shen ldquoResearch on lie group kernellearning algorithmrdquo Journal of Frontiers of Compter Science andTechnology vol 6 pp 1026ndash1038 2012

[22] V N Vapnik and A Lerner ldquoPattern recognition using general-ized portrait methodrdquo Automation and Remote Control vol 24pp 774ndash780 1963

[23] V Vapnik The Nature of Statistical Learning Theory SpringerNew York NY USA 2000

[24] D TaxOne-class classification [PhD thesis] Delft University ofTechnology 2001

12 Journal of Electrical and Computer Engineering

[25] D M Tax and R P Duin ldquoData domain description usingsupport vectorsrdquo in Proceedings of the European Symposiumon Artificial Neural Networks Computational Intelligence andMachine Learning (ESANN rsquo99) vol 99 pp 251ndash256 1999

[26] B Scholkopf J C Platt J Shawe-Taylor A J Smola and RC Williamson ldquoEstimating the support of a high-dimensionaldistributionrdquo Neural Computation vol 13 no 7 pp 1443ndash14712001

[27] Z Noumir P Honeine and C Richard ldquoOnline one-classmachines based on the coherence criterionrdquo in Proceedings ofthe 20th European Signal Processing Conference (EUSIPCO rsquo12)pp 664ndash668 2012

[28] Z Noumir P Honeine and C Richard ldquoOne-class machinesbased on the coherence criterionrdquo in Proceedings of the IEEEStatistical Signal Processing Workshop (SSP rsquo12) pp 600ndash6032012

[29] P Honeine ldquoOnline kernel principal component analysis areducedorder modelrdquo IEEE Transactions on Pattern Analysisand Machine Intelligence vol 34 no 9 pp 1814ndash1826 2012

[30] C Richard J C M Bermudez and P Honeine ldquoOnlineprediction of time series data with kernelsrdquo IEEE Transactionson Signal Processing vol 57 no 3 pp 1058ndash1067 2009

[31] UMN ldquoUnusual crowd activity dataset of university of Min-nesotardquo 2006 httpmhacsumneduproj eventsshtml

[32] J A Hanley and B J McNeil ldquoThe meaning and use of thearea under a receiver operating characteristic (ROC) curverdquoRadiology vol 143 no 1 pp 29ndash36 1982

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

6 Journal of Electrical and Computer Engineering

41 Strategy 1 In Strategy 1 the dictionary is updated merelythrough the training period

Step 1 The first step is calculating the covariance matrixdescriptor of training frames based on the image intensity andthe optical flow This step can be generalized as

(1198681OP1) (1198682OP2) (119868

119899OP119899) 997888rarr C

1C2 C

119899

(21)

where (I1OP1) (I2OP2) (119868

119899OP119899) are the image

intensity and the corresponding optical flow of the 1st to 119899thframe C

1C2 C

119899 are the covariance matrix descriptors

Step 2 The second step consists of applying one-class SVMon the small subset of extracted descriptor of the trainingnormal frames to obtain the support vectors Consider asubset C

119894119898

119894=1 1 le 119898 ≪ 119899 of data selected from the full

training sample set C119894119899

119894=1 without loss of generality and

assume that the first 119898 examples are chosen This set of 119898examples is called dictionary CD

C1C2 C

119898 1 le 119898 ≪ 119899

SVM997888rarr support vector 119878119901

1 1198781199012 119878119901

119900

(22)

where the set C1C2 C

119898 is the first119898 covariancematrix

descriptors of the training frames it is the original dictionaryCD In one-class SVM the majority of the training samplesdo not contribute to the definition of the decision functionThe entries of a monitory subset of the training samples1198781199011 1198781199012 119878119901

119900 119900 le 119898 are support vectors contributing

to the definition of the decision function

Step 3 After learning the dictionary CD which includesthe first 119898 1 le 119898 ≪ 119899 samples the training samplesC119898+1

C119898+2

C119899 are learned online via the technique

described in Section 3 This step can be generalized as

CDC119896 119898 lt 119896 le 119899SVM997888rarr support vector 119878119901

1 1198781199012 119878119901

119901

119900 le 119901 le 119899

CD = CD cup C119896 if 120598

119905ge 1205830

(23)

where CD is the dictionary obtained through Step 2 C119896is a

new sample in the remaining training dataset According tothe criterion introduced in Section 3 if the new sample C

119896

satisfies the dictionary updated condition it will be includedinto the dictionary CD

Step 4 Based on the dictionary and the classifier obtainedfrom the training frames the incoming frame sample C

119899+119897is

Table 2 AUC of abnormal event detection method of differentfeatures 119865 via original SVDD which learns training samples offlineStrategy 1 online one-class SVM and Strategy 2 online one-classSVMThe biggest value of each method is shown in bold

Features Area under ROCLawn Indoor Plaza

Training samples are learned offline1198651(6 times 6119889119906) 09426 08351 09323

1198652(6 times 6119889V) 09400 08358 09321

1198653(8 times 8) 09440 08375 09359

1198654(12 times 12) 09591 08440 09580

1198655(17 times 17) 09567 08649 09649

Strategy 11198651(6 times 6119889119906) 09399 08328 09343

1198652(6 times 6119889V) 09390 08355 09366

1198653(8 times 8) 09418 08377 09411

1198654(12 times 12) 09581 08457 09573

1198655(17 times 17) 09551 08628 09632

Strategy 21198651(6 times 6119889119906) 09427 08237 09288

1198652(6 times 6119889V) 09370 08241 09283

1198653(8 times 8) 09430 08274 09312

1198654(12 times 12) 09605 08331 09505

1198655(17 times 17) 09601 08495 09746

classifiedTheworkflow of Strategy 1 is shown in Figure 4 anddescribed by the following equation

119891 (C119899+119897) = sgn(1198772 minus

119899

sum

119894119895=1

120572119894120572119895120581 (C119894C119895)

+ 2sum

119894

120572119894120581 (C119894C119899+119897) minus 120581 (C

119899+119897C119899+119897))

= 1 119891 (C

119899+119897) ge 0

minus1 119891 (C119899+119897) lt 0

(24)

whereC119899+119897

is the covariance matrix descriptor of the (119899+ 119897)thframe needed to be classified and C

119894and C

119895are the samples

of the dictionary CD ldquo1rdquo corresponds to the normal frameldquominus1rdquo corresponds to the abnormal frame

42 Strategy 2 In this strategy the dictionary is updatedthrough both training and testing periodsThe feature extrac-tion step (Step 1) and the online training steps (Steps 2 and 3)are the same as the ones presented in Strategy 1 The testingstep is different The new coming datum which is detectedas normal but satisfies dictionary update condition shouldbe included into CD The dictionary is needed to be updatedthrough the testing period to include new samples

Step 4 Strategy 2 If the incoming frame sample C119899+119897

isclassified as normal (119891(C

119899+119897) = 1) the data is checked by

the criterion described in Section 3 When the data satisfies

Journal of Electrical and Computer Engineering 7

(a) Normal lawn scene (b) Abnormal lawn scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC lawn offline

094092

09088086084082

080 005 01 015 02

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

094092

09088086084082

080 005 01 015 02

ROC lawn Strategy 1

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 5 Detection results of the lawn scene (a) The detection result of one normal frame (b) The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the lawn scene results via one-class SVM All the training samples are learned together offlineThe biggest AUC value is 09591 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUC value is09581

the dictionary update criterion this testing sample will beincluded into the dictionary This step can be generalized bythe following equation

119891 (C119899+119897) = sgn(1198772 minus

119899

sum

119894119895=1

120572119894120572119895120581 (C119894C119895)

+ 2sum

119894

120572119894120581 (C119894C119899+119897) minus 120581 (C

119899+119897C119899+119897))

=

1 119891 (C119899+119897) ge 0

120598119905ge 1205830997888rarr CD = CD cup C

119899+119897

120598119905lt 1205830997888rarr CD = CD

minus1 119891 (C119899+119897) lt 0

(25)

5 Abnormal Detection Result

This section presents the results of experiments conducted toanalyze the performance of the proposedmethod A compet-itive performance through both Strategy 1 and Strategy 2 ofUMN [31] dataset is presented

51 Abnormal Visual Events Detection Strategy 1 The resultsof the proposed abnormal events detection method viaStrategy 1 online one-class SVM of UMN [31] dataset areshown below

The UMN dataset includes eleven video sequences ofthree different scenes (the lawn indoor and plaza) ofcrowded escape events The normal samples for training or

8 Journal of Electrical and Computer Engineering

(a) Normal indoor scene (b) Abnormal indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC indoor offline

085

08

075

07

065

0 01 02 03 04

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC indoor Strategy 1

085

08

075

07

065

0 01 02 03 04

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 6 Detection results of the indoor scene (a)The detection result of one normal frame (b)The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the indoor scene results via one-class SVM All the training samples are learned togetheroffline The biggest AUC value is 08649 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUCvalue is 08628

Table 3 The comparison of our proposed method with the state-of-the-art methods for the abnormal event detection in the UMNdataset

Method Area under ROCLawn Indoor Plaza

Social force [16] 096Optical flow [16] 084NN [17] 093SRC [17] 0995 0975 0964STCOG [18] 09362 07759 09661COV online (ours) 09605 08628 09746

for normal testing are the frames where the people are walk-ing in different directions The samples for abnormal testingare the frames where people are running The detectionresults of the lawn scene indoor scene and plaza scene areshown in Figures 5 6 and 7 respectively Gaussian kernelfor the Lie Group is used in these three scenes Differentvalues of120590 and penalty factor119862 are chosen the area under theROC curve is shown as a function of these parameters [32]The results show that taking covariance matrix as descriptorcan obtain satisfactory performance for abnormal detectionAnd also training the samples online can obtain similarlydetection performance as training all the samples offline

Journal of Electrical and Computer Engineering 9

(a) Normal plaza scene (b) Abnormal plaza scene

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

ROC plaza offline

False positive

True

pos

itive

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

False positive

True

pos

itive

ROC plaza Strategy 1

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 7 Detection results of the plaza scene (a) The detection result of one normal frame (b) The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the plaza scene results via one-class SVM All the training samples are learned together offlineThe biggest AUC value is 09649 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUC value is09632

Online one-class SVM is appropriate to detect abnormalvisual events There are 1431 frames in the lawn scene and480 normal frames are used for training In the offline strat-egy all the 480 frames covariancematrices should be saved inthememory In Strategy 1 100 frames covariancematrices areconsidered as the dictionary firstlyWhen119865

5-17times17 feature is

adopted to construct the covariance descriptor the varianceof Gaussian kernel is 120590 = 1 the preset threshold of thecriterion is 120583

0= 05 the dictionary size increases from 100

to 101 and the maximum accuracy of the detection resultsis 9169 In the indoor scene there are 2975 normal framesand 1057 abnormal frames In the plaza scene there are 1831normal frames and 286 abnormal framesTheprocesses of theexperiments are similar to the ones of the lawn scene Whenfeature vector is 119865

5-17 times 17 120590 = 1 120583

0= 05 the dictionary

size of these two scenes remain 100The online strategy keepsthe memory size almost unchanged when the size of trainingdataset increases

52 Abnormal Visual Events Detection Strategy 2 The resultsof the abnormal event detection method via Strategy 2 ofUMNdataset are shown as follows In the experiment processof the lawn scene 100 normal samples from the trainingsamples are learnt firstly and then the other 380 trainingdata are learnt online one by one After these two trainingsteps we can obtain the basic dictionary from the trainingsamples and also the classifier In the following testing stepthe dictionary is updated if the sample satisfies the dictionaryupdate criterion When a new sample is coming it is firstlydetected by the previous classifier If it is classified as anomaly

10 Journal of Electrical and Computer Engineering

094092

09088086084082

08

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC lawn Strategy 2

F1

F2

F3

F4

F5

(a) ROC Strategy 2 lawn scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

085

08

075

07

065

0 01 02 03 04

ROC indoor Strategy 2

F1

F2

F3

F4

F5

(b) ROC Strategy 2 indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC plaza Strategy 2

F1

F2

F3

F4

F5

098096094092

09088086084082

08

(c) ROC Strategy 2 plaza scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC train data learnt offline

LawnPlazaIndoor

(d) ROC train offline

Figure 8 ROC curve of UMN dataset (a) ROC curve of different features 119865 results via Strategy 2 of the lawn sceneThe biggest AUC value is09605 (b) Strategy 2 results of the indoor scene The biggest AUC value is 08495 (c) Strategy 2 results of the plaza scene The biggest AUCvalue is 09746 (d)The ROC curve of best performance of the lawn indoor and plaza scene when the training samples are learnt offlineThebiggest AUC values of the lawn indoor and plaza are 09591 08649 and 09649

the dictionary and the classifier are not changed Otherwiseif the sample is classified as a normal one the sparse criterionintroduced in Section 3 is used to check the correlationbetween the earlier dictionary and this new datum It willbe included into the dictionary when it satisfied the updatecondition The dictionary will be updated through the wholetesting period The other two scenes the indoor and plazascene are handled by the same methods When 119865

5-17 times 17

feature is adopted the variance of the Gaussian kernel is120590 = 1 and the preset threshold of the criterion is 120583

0= 05

and the dictionary size of the lawn indoor and plaza scene

is increased from 100 to 106 102 and 102 respectively TheROC curve of detection results of these three scenes is shownin Figures 8(a) 8(b) and 8(c) Besides the merit of savingmemory of Strategy 1 Strategy 2 also has the advantage ofadaptation to the long duration sequence

The results performances of offline strategy Strategy 1and Strategy 2 are shown in Table 2 The performances ofthese two strategies results are similar to that of the resultswhen all training samples are learnt together When 119865

4(12 times

12) or1198655(17times17) are chosen as the features to formcovariance

matrix descriptor the results have the best performance

Journal of Electrical and Computer Engineering 11

These two features are more abundant to include movementand intensity information

The result performances of the covariancematrix descrip-tor based online one-class SVM method and the state-of-the-art methods are shown in Table 3 The covariance matrixbased online abnormal frame detection method obtainscompetitive performance In general our method is betterthan others except sparse reconstruction cost (SRC) [17]in lawn scene and indoor scene In that paper multiscaleHOF is taken as a feature and a testing sample is classi-fied by its sparse reconstructor cost through a weightedlinear reconstruction of the overcomplete normal basis setBut computation of the HOF might takes more time thancalculating covariance By adopting the integral image [15]the covariance matrix descriptor of the subimage can becomputed conveniently So the covariance descriptor can beappropriately used to analyze the partial movement

6 Conclusions

A method for the abnormal event detection of the frameis proposed The method consists of covariance matrixdescriptor encoding the movement features and the onlinenonlinear one-class SVM classification method We havedeveloped two nonlinear one-class SVM based abnormalevent detection techniques that update the normal modelsof the surveillance video data in an online framework Theproposed algorithm has been tested on a video datasetyielding successful results to detect abnormal events

Acknowledgments

This work is partially supported by China ScholarshipCouncil of Chinese Government SURECAP CPER Projectfunded by Region Champagne-Ardenne and CAPSECCRCA FEDER platform

References

[1] F Jiang J Yuan S A Tsaftaris and A K Katsaggelos ldquoAnoma-lous video event detection using spatiotemporal contextrdquo Com-puter Vision and Image Understanding vol 115 no 3 pp 323ndash333 2011

[2] C Piciarelli C Micheloni and G L Foresti ldquoTrajectory-basedanomalous event detectionrdquo IEEE Transactions on Circuits andSystems for Video Technology vol 18 no 11 pp 1544ndash1554 2008

[3] C Piciarelli and G L Foresti ldquoOn-line trajectory clustering foranomalous events detectionrdquo Pattern Recognition Letters vol27 no 15 pp 1835ndash1842 2006

[4] T Xiang and S Gong ldquoIncremental and adaptive abnormalbehaviour detectionrdquo Computer Vision and Image Understand-ing vol 111 no 1 pp 59ndash73 2008

[5] S Gong and T Xiang ldquoRecognition of group activities usingdynamic probabilistic networksrdquo in Proceedings of the 9th IEEEInternational Conference on Computer Vision (ICCV rsquo03) pp742ndash749 October 2003

[6] J KimandKGrauman ldquoObserve locally infer globally a space-time MRF for detecting abnormal activities with incrementalupdatesrdquo in Proceedings of the IEEE Conference on Computer

Vision and Pattern Recognition (CVPR rsquo09) pp 2921ndash2928Miami Fla USA June 2009

[7] A Adam E Rivlin I Shimshoni and D Reinitz ldquoRobustreal-time unusual event detection using multiple fixed-locationmonitorsrdquo IEEE Transactions on Pattern Analysis and MachineIntelligence vol 30 no 3 pp 555ndash560 2008

[8] O Boiman andM Irani ldquoDetecting irregularities in images andin videordquo International Journal of Computer Vision vol 74 no1 pp 17ndash31 2007

[9] X Zhu Z Liu and J Zhang ldquoHuman activity clustering foronline anomaly detectionrdquo Journal of Computers vol 6 no 6pp 1071ndash1079 2011

[10] X Zhu and Z Liu ldquoHuman behavior clustering for anomalydetectionrdquo Frontiers of Computer Science in China vol 5 no3 pp 279ndash289 2011

[11] Y Benezeth P M Jodoin and V Saligrama ldquoAbnormalitydetection using low-level co-occurring eventsrdquo Pattern Recog-nition Letters vol 32 no 3 pp 423ndash431 2011

[12] Y Benezeth P M Jodoin V Saligrama and C RosenbergerldquoAbnormal events detection based on spatio-temporal co-occurencesrdquo in Proceedings of the IEEE Conference on ComputerVision and Pattern Recognition (CVPR rsquo09) pp 2458ndash2465Miami Fla USA June 2009

[13] A Mittal A Monnet and N Paragios ldquoScene modeling andchange detection in dynamic scenes a subspace approachrdquoComputer Vision and Image Understanding vol 113 no 1 pp63ndash79 2009

[14] B K P Horn and B G Schunck ldquoDetermining optical flowrdquoArtificial Intelligence vol 17 no 1ndash3 pp 185ndash203 1981

[15] O Tuzel F Porikli and P Meer ldquoRegion covariance afast descriptor for detection and classificationrdquo in ComputerVisionmdashECCV 2006 vol 3952 of Lecture Notes in ComputerScience pp 589ndash600 Springer New York NY USA 2006

[16] RMehran A Oyama andM Shah ldquoAbnormal crowd behaviordetection using social force modelrdquo in Proceedings of the IEEEConference on Computer Vision and Pattern Recognition (CVPRrsquo09) pp 935ndash942 Miami Fla USA June 2009

[17] Y Cong J Yuan and J Liu ldquoSparse reconstruction cost forabnormal event detectionrdquo in Proceedings of the IEEE Confer-ence on Computer Vision and Pattern Recognition (CVPR rsquo11)pp 3449ndash3456 Providence RI USA June 2011

[18] Y Shi Y Gao and R Wang ldquoReal-time abnormal eventdetection in complicated scenesrdquo in Proceedings of the 20thInternational Conference on Pattern Recognition (ICPR rsquo10) pp3653ndash3656 Istanbul Turkey August 2010

[19] B Scholkopf and A J Smola Learning with Kernels SupportVector Machines Regularization Optimization and BeyondMIT Press Boston Mass USA 2002

[20] B Hall Lie Groups Lie Algebras And Representations AnElementary Introduction vol 222 Springer New York NYUSA 2003

[21] C Gao F Li and C Shen ldquoResearch on lie group kernellearning algorithmrdquo Journal of Frontiers of Compter Science andTechnology vol 6 pp 1026ndash1038 2012

[22] V N Vapnik and A Lerner ldquoPattern recognition using general-ized portrait methodrdquo Automation and Remote Control vol 24pp 774ndash780 1963

[23] V Vapnik The Nature of Statistical Learning Theory SpringerNew York NY USA 2000

[24] D TaxOne-class classification [PhD thesis] Delft University ofTechnology 2001

12 Journal of Electrical and Computer Engineering

[25] D M Tax and R P Duin ldquoData domain description usingsupport vectorsrdquo in Proceedings of the European Symposiumon Artificial Neural Networks Computational Intelligence andMachine Learning (ESANN rsquo99) vol 99 pp 251ndash256 1999

[26] B Scholkopf J C Platt J Shawe-Taylor A J Smola and RC Williamson ldquoEstimating the support of a high-dimensionaldistributionrdquo Neural Computation vol 13 no 7 pp 1443ndash14712001

[27] Z Noumir P Honeine and C Richard ldquoOnline one-classmachines based on the coherence criterionrdquo in Proceedings ofthe 20th European Signal Processing Conference (EUSIPCO rsquo12)pp 664ndash668 2012

[28] Z Noumir P Honeine and C Richard ldquoOne-class machinesbased on the coherence criterionrdquo in Proceedings of the IEEEStatistical Signal Processing Workshop (SSP rsquo12) pp 600ndash6032012

[29] P Honeine ldquoOnline kernel principal component analysis areducedorder modelrdquo IEEE Transactions on Pattern Analysisand Machine Intelligence vol 34 no 9 pp 1814ndash1826 2012

[30] C Richard J C M Bermudez and P Honeine ldquoOnlineprediction of time series data with kernelsrdquo IEEE Transactionson Signal Processing vol 57 no 3 pp 1058ndash1067 2009

[31] UMN ldquoUnusual crowd activity dataset of university of Min-nesotardquo 2006 httpmhacsumneduproj eventsshtml

[32] J A Hanley and B J McNeil ldquoThe meaning and use of thearea under a receiver operating characteristic (ROC) curverdquoRadiology vol 143 no 1 pp 29ndash36 1982

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Journal of Electrical and Computer Engineering 7

(a) Normal lawn scene (b) Abnormal lawn scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC lawn offline

094092

09088086084082

080 005 01 015 02

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

094092

09088086084082

080 005 01 015 02

ROC lawn Strategy 1

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 5 Detection results of the lawn scene (a) The detection result of one normal frame (b) The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the lawn scene results via one-class SVM All the training samples are learned together offlineThe biggest AUC value is 09591 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUC value is09581

the dictionary update criterion this testing sample will beincluded into the dictionary This step can be generalized bythe following equation

119891 (C119899+119897) = sgn(1198772 minus

119899

sum

119894119895=1

120572119894120572119895120581 (C119894C119895)

+ 2sum

119894

120572119894120581 (C119894C119899+119897) minus 120581 (C

119899+119897C119899+119897))

=

1 119891 (C119899+119897) ge 0

120598119905ge 1205830997888rarr CD = CD cup C

119899+119897

120598119905lt 1205830997888rarr CD = CD

minus1 119891 (C119899+119897) lt 0

(25)

5 Abnormal Detection Result

This section presents the results of experiments conducted toanalyze the performance of the proposedmethod A compet-itive performance through both Strategy 1 and Strategy 2 ofUMN [31] dataset is presented

51 Abnormal Visual Events Detection Strategy 1 The resultsof the proposed abnormal events detection method viaStrategy 1 online one-class SVM of UMN [31] dataset areshown below

The UMN dataset includes eleven video sequences ofthree different scenes (the lawn indoor and plaza) ofcrowded escape events The normal samples for training or

8 Journal of Electrical and Computer Engineering

(a) Normal indoor scene (b) Abnormal indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC indoor offline

085

08

075

07

065

0 01 02 03 04

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC indoor Strategy 1

085

08

075

07

065

0 01 02 03 04

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 6 Detection results of the indoor scene (a)The detection result of one normal frame (b)The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the indoor scene results via one-class SVM All the training samples are learned togetheroffline The biggest AUC value is 08649 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUCvalue is 08628

Table 3 The comparison of our proposed method with the state-of-the-art methods for the abnormal event detection in the UMNdataset

Method Area under ROCLawn Indoor Plaza

Social force [16] 096Optical flow [16] 084NN [17] 093SRC [17] 0995 0975 0964STCOG [18] 09362 07759 09661COV online (ours) 09605 08628 09746

for normal testing are the frames where the people are walk-ing in different directions The samples for abnormal testingare the frames where people are running The detectionresults of the lawn scene indoor scene and plaza scene areshown in Figures 5 6 and 7 respectively Gaussian kernelfor the Lie Group is used in these three scenes Differentvalues of120590 and penalty factor119862 are chosen the area under theROC curve is shown as a function of these parameters [32]The results show that taking covariance matrix as descriptorcan obtain satisfactory performance for abnormal detectionAnd also training the samples online can obtain similarlydetection performance as training all the samples offline

Journal of Electrical and Computer Engineering 9

(a) Normal plaza scene (b) Abnormal plaza scene

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

ROC plaza offline

False positive

True

pos

itive

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

False positive

True

pos

itive

ROC plaza Strategy 1

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 7 Detection results of the plaza scene (a) The detection result of one normal frame (b) The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the plaza scene results via one-class SVM All the training samples are learned together offlineThe biggest AUC value is 09649 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUC value is09632

Online one-class SVM is appropriate to detect abnormalvisual events There are 1431 frames in the lawn scene and480 normal frames are used for training In the offline strat-egy all the 480 frames covariancematrices should be saved inthememory In Strategy 1 100 frames covariancematrices areconsidered as the dictionary firstlyWhen119865

5-17times17 feature is

adopted to construct the covariance descriptor the varianceof Gaussian kernel is 120590 = 1 the preset threshold of thecriterion is 120583

0= 05 the dictionary size increases from 100

to 101 and the maximum accuracy of the detection resultsis 9169 In the indoor scene there are 2975 normal framesand 1057 abnormal frames In the plaza scene there are 1831normal frames and 286 abnormal framesTheprocesses of theexperiments are similar to the ones of the lawn scene Whenfeature vector is 119865

5-17 times 17 120590 = 1 120583

0= 05 the dictionary

size of these two scenes remain 100The online strategy keepsthe memory size almost unchanged when the size of trainingdataset increases

52 Abnormal Visual Events Detection Strategy 2 The resultsof the abnormal event detection method via Strategy 2 ofUMNdataset are shown as follows In the experiment processof the lawn scene 100 normal samples from the trainingsamples are learnt firstly and then the other 380 trainingdata are learnt online one by one After these two trainingsteps we can obtain the basic dictionary from the trainingsamples and also the classifier In the following testing stepthe dictionary is updated if the sample satisfies the dictionaryupdate criterion When a new sample is coming it is firstlydetected by the previous classifier If it is classified as anomaly

10 Journal of Electrical and Computer Engineering

094092

09088086084082

08

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC lawn Strategy 2

F1

F2

F3

F4

F5

(a) ROC Strategy 2 lawn scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

085

08

075

07

065

0 01 02 03 04

ROC indoor Strategy 2

F1

F2

F3

F4

F5

(b) ROC Strategy 2 indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC plaza Strategy 2

F1

F2

F3

F4

F5

098096094092

09088086084082

08

(c) ROC Strategy 2 plaza scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC train data learnt offline

LawnPlazaIndoor

(d) ROC train offline

Figure 8 ROC curve of UMN dataset (a) ROC curve of different features 119865 results via Strategy 2 of the lawn sceneThe biggest AUC value is09605 (b) Strategy 2 results of the indoor scene The biggest AUC value is 08495 (c) Strategy 2 results of the plaza scene The biggest AUCvalue is 09746 (d)The ROC curve of best performance of the lawn indoor and plaza scene when the training samples are learnt offlineThebiggest AUC values of the lawn indoor and plaza are 09591 08649 and 09649

the dictionary and the classifier are not changed Otherwiseif the sample is classified as a normal one the sparse criterionintroduced in Section 3 is used to check the correlationbetween the earlier dictionary and this new datum It willbe included into the dictionary when it satisfied the updatecondition The dictionary will be updated through the wholetesting period The other two scenes the indoor and plazascene are handled by the same methods When 119865

5-17 times 17

feature is adopted the variance of the Gaussian kernel is120590 = 1 and the preset threshold of the criterion is 120583

0= 05

and the dictionary size of the lawn indoor and plaza scene

is increased from 100 to 106 102 and 102 respectively TheROC curve of detection results of these three scenes is shownin Figures 8(a) 8(b) and 8(c) Besides the merit of savingmemory of Strategy 1 Strategy 2 also has the advantage ofadaptation to the long duration sequence

The results performances of offline strategy Strategy 1and Strategy 2 are shown in Table 2 The performances ofthese two strategies results are similar to that of the resultswhen all training samples are learnt together When 119865

4(12 times

12) or1198655(17times17) are chosen as the features to formcovariance

matrix descriptor the results have the best performance

Journal of Electrical and Computer Engineering 11

These two features are more abundant to include movementand intensity information

The result performances of the covariancematrix descrip-tor based online one-class SVM method and the state-of-the-art methods are shown in Table 3 The covariance matrixbased online abnormal frame detection method obtainscompetitive performance In general our method is betterthan others except sparse reconstruction cost (SRC) [17]in lawn scene and indoor scene In that paper multiscaleHOF is taken as a feature and a testing sample is classi-fied by its sparse reconstructor cost through a weightedlinear reconstruction of the overcomplete normal basis setBut computation of the HOF might takes more time thancalculating covariance By adopting the integral image [15]the covariance matrix descriptor of the subimage can becomputed conveniently So the covariance descriptor can beappropriately used to analyze the partial movement

6 Conclusions

A method for the abnormal event detection of the frameis proposed The method consists of covariance matrixdescriptor encoding the movement features and the onlinenonlinear one-class SVM classification method We havedeveloped two nonlinear one-class SVM based abnormalevent detection techniques that update the normal modelsof the surveillance video data in an online framework Theproposed algorithm has been tested on a video datasetyielding successful results to detect abnormal events

Acknowledgments

This work is partially supported by China ScholarshipCouncil of Chinese Government SURECAP CPER Projectfunded by Region Champagne-Ardenne and CAPSECCRCA FEDER platform

References

[1] F Jiang J Yuan S A Tsaftaris and A K Katsaggelos ldquoAnoma-lous video event detection using spatiotemporal contextrdquo Com-puter Vision and Image Understanding vol 115 no 3 pp 323ndash333 2011

[2] C Piciarelli C Micheloni and G L Foresti ldquoTrajectory-basedanomalous event detectionrdquo IEEE Transactions on Circuits andSystems for Video Technology vol 18 no 11 pp 1544ndash1554 2008

[3] C Piciarelli and G L Foresti ldquoOn-line trajectory clustering foranomalous events detectionrdquo Pattern Recognition Letters vol27 no 15 pp 1835ndash1842 2006

[4] T Xiang and S Gong ldquoIncremental and adaptive abnormalbehaviour detectionrdquo Computer Vision and Image Understand-ing vol 111 no 1 pp 59ndash73 2008

[5] S Gong and T Xiang ldquoRecognition of group activities usingdynamic probabilistic networksrdquo in Proceedings of the 9th IEEEInternational Conference on Computer Vision (ICCV rsquo03) pp742ndash749 October 2003

[6] J KimandKGrauman ldquoObserve locally infer globally a space-time MRF for detecting abnormal activities with incrementalupdatesrdquo in Proceedings of the IEEE Conference on Computer

Vision and Pattern Recognition (CVPR rsquo09) pp 2921ndash2928Miami Fla USA June 2009

[7] A Adam E Rivlin I Shimshoni and D Reinitz ldquoRobustreal-time unusual event detection using multiple fixed-locationmonitorsrdquo IEEE Transactions on Pattern Analysis and MachineIntelligence vol 30 no 3 pp 555ndash560 2008

[8] O Boiman andM Irani ldquoDetecting irregularities in images andin videordquo International Journal of Computer Vision vol 74 no1 pp 17ndash31 2007

[9] X Zhu Z Liu and J Zhang ldquoHuman activity clustering foronline anomaly detectionrdquo Journal of Computers vol 6 no 6pp 1071ndash1079 2011

[10] X Zhu and Z Liu ldquoHuman behavior clustering for anomalydetectionrdquo Frontiers of Computer Science in China vol 5 no3 pp 279ndash289 2011

[11] Y Benezeth P M Jodoin and V Saligrama ldquoAbnormalitydetection using low-level co-occurring eventsrdquo Pattern Recog-nition Letters vol 32 no 3 pp 423ndash431 2011

[12] Y Benezeth P M Jodoin V Saligrama and C RosenbergerldquoAbnormal events detection based on spatio-temporal co-occurencesrdquo in Proceedings of the IEEE Conference on ComputerVision and Pattern Recognition (CVPR rsquo09) pp 2458ndash2465Miami Fla USA June 2009

[13] A Mittal A Monnet and N Paragios ldquoScene modeling andchange detection in dynamic scenes a subspace approachrdquoComputer Vision and Image Understanding vol 113 no 1 pp63ndash79 2009

[14] B K P Horn and B G Schunck ldquoDetermining optical flowrdquoArtificial Intelligence vol 17 no 1ndash3 pp 185ndash203 1981

[15] O Tuzel F Porikli and P Meer ldquoRegion covariance afast descriptor for detection and classificationrdquo in ComputerVisionmdashECCV 2006 vol 3952 of Lecture Notes in ComputerScience pp 589ndash600 Springer New York NY USA 2006

[16] RMehran A Oyama andM Shah ldquoAbnormal crowd behaviordetection using social force modelrdquo in Proceedings of the IEEEConference on Computer Vision and Pattern Recognition (CVPRrsquo09) pp 935ndash942 Miami Fla USA June 2009

[17] Y Cong J Yuan and J Liu ldquoSparse reconstruction cost forabnormal event detectionrdquo in Proceedings of the IEEE Confer-ence on Computer Vision and Pattern Recognition (CVPR rsquo11)pp 3449ndash3456 Providence RI USA June 2011

[18] Y Shi Y Gao and R Wang ldquoReal-time abnormal eventdetection in complicated scenesrdquo in Proceedings of the 20thInternational Conference on Pattern Recognition (ICPR rsquo10) pp3653ndash3656 Istanbul Turkey August 2010

[19] B Scholkopf and A J Smola Learning with Kernels SupportVector Machines Regularization Optimization and BeyondMIT Press Boston Mass USA 2002

[20] B Hall Lie Groups Lie Algebras And Representations AnElementary Introduction vol 222 Springer New York NYUSA 2003

[21] C Gao F Li and C Shen ldquoResearch on lie group kernellearning algorithmrdquo Journal of Frontiers of Compter Science andTechnology vol 6 pp 1026ndash1038 2012

[22] V N Vapnik and A Lerner ldquoPattern recognition using general-ized portrait methodrdquo Automation and Remote Control vol 24pp 774ndash780 1963

[23] V Vapnik The Nature of Statistical Learning Theory SpringerNew York NY USA 2000

[24] D TaxOne-class classification [PhD thesis] Delft University ofTechnology 2001

12 Journal of Electrical and Computer Engineering

[25] D M Tax and R P Duin ldquoData domain description usingsupport vectorsrdquo in Proceedings of the European Symposiumon Artificial Neural Networks Computational Intelligence andMachine Learning (ESANN rsquo99) vol 99 pp 251ndash256 1999

[26] B Scholkopf J C Platt J Shawe-Taylor A J Smola and RC Williamson ldquoEstimating the support of a high-dimensionaldistributionrdquo Neural Computation vol 13 no 7 pp 1443ndash14712001

[27] Z Noumir P Honeine and C Richard ldquoOnline one-classmachines based on the coherence criterionrdquo in Proceedings ofthe 20th European Signal Processing Conference (EUSIPCO rsquo12)pp 664ndash668 2012

[28] Z Noumir P Honeine and C Richard ldquoOne-class machinesbased on the coherence criterionrdquo in Proceedings of the IEEEStatistical Signal Processing Workshop (SSP rsquo12) pp 600ndash6032012

[29] P Honeine ldquoOnline kernel principal component analysis areducedorder modelrdquo IEEE Transactions on Pattern Analysisand Machine Intelligence vol 34 no 9 pp 1814ndash1826 2012

[30] C Richard J C M Bermudez and P Honeine ldquoOnlineprediction of time series data with kernelsrdquo IEEE Transactionson Signal Processing vol 57 no 3 pp 1058ndash1067 2009

[31] UMN ldquoUnusual crowd activity dataset of university of Min-nesotardquo 2006 httpmhacsumneduproj eventsshtml

[32] J A Hanley and B J McNeil ldquoThe meaning and use of thearea under a receiver operating characteristic (ROC) curverdquoRadiology vol 143 no 1 pp 29ndash36 1982

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

8 Journal of Electrical and Computer Engineering

(a) Normal indoor scene (b) Abnormal indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC indoor offline

085

08

075

07

065

0 01 02 03 04

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC indoor Strategy 1

085

08

075

07

065

0 01 02 03 04

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 6 Detection results of the indoor scene (a)The detection result of one normal frame (b)The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the indoor scene results via one-class SVM All the training samples are learned togetheroffline The biggest AUC value is 08649 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUCvalue is 08628

Table 3 The comparison of our proposed method with the state-of-the-art methods for the abnormal event detection in the UMNdataset

Method Area under ROCLawn Indoor Plaza

Social force [16] 096Optical flow [16] 084NN [17] 093SRC [17] 0995 0975 0964STCOG [18] 09362 07759 09661COV online (ours) 09605 08628 09746

for normal testing are the frames where the people are walk-ing in different directions The samples for abnormal testingare the frames where people are running The detectionresults of the lawn scene indoor scene and plaza scene areshown in Figures 5 6 and 7 respectively Gaussian kernelfor the Lie Group is used in these three scenes Differentvalues of120590 and penalty factor119862 are chosen the area under theROC curve is shown as a function of these parameters [32]The results show that taking covariance matrix as descriptorcan obtain satisfactory performance for abnormal detectionAnd also training the samples online can obtain similarlydetection performance as training all the samples offline

Journal of Electrical and Computer Engineering 9

(a) Normal plaza scene (b) Abnormal plaza scene

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

ROC plaza offline

False positive

True

pos

itive

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

False positive

True

pos

itive

ROC plaza Strategy 1

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 7 Detection results of the plaza scene (a) The detection result of one normal frame (b) The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the plaza scene results via one-class SVM All the training samples are learned together offlineThe biggest AUC value is 09649 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUC value is09632

Online one-class SVM is appropriate to detect abnormalvisual events There are 1431 frames in the lawn scene and480 normal frames are used for training In the offline strat-egy all the 480 frames covariancematrices should be saved inthememory In Strategy 1 100 frames covariancematrices areconsidered as the dictionary firstlyWhen119865

5-17times17 feature is

adopted to construct the covariance descriptor the varianceof Gaussian kernel is 120590 = 1 the preset threshold of thecriterion is 120583

0= 05 the dictionary size increases from 100

to 101 and the maximum accuracy of the detection resultsis 9169 In the indoor scene there are 2975 normal framesand 1057 abnormal frames In the plaza scene there are 1831normal frames and 286 abnormal framesTheprocesses of theexperiments are similar to the ones of the lawn scene Whenfeature vector is 119865

5-17 times 17 120590 = 1 120583

0= 05 the dictionary

size of these two scenes remain 100The online strategy keepsthe memory size almost unchanged when the size of trainingdataset increases

52 Abnormal Visual Events Detection Strategy 2 The resultsof the abnormal event detection method via Strategy 2 ofUMNdataset are shown as follows In the experiment processof the lawn scene 100 normal samples from the trainingsamples are learnt firstly and then the other 380 trainingdata are learnt online one by one After these two trainingsteps we can obtain the basic dictionary from the trainingsamples and also the classifier In the following testing stepthe dictionary is updated if the sample satisfies the dictionaryupdate criterion When a new sample is coming it is firstlydetected by the previous classifier If it is classified as anomaly

10 Journal of Electrical and Computer Engineering

094092

09088086084082

08

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC lawn Strategy 2

F1

F2

F3

F4

F5

(a) ROC Strategy 2 lawn scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

085

08

075

07

065

0 01 02 03 04

ROC indoor Strategy 2

F1

F2

F3

F4

F5

(b) ROC Strategy 2 indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC plaza Strategy 2

F1

F2

F3

F4

F5

098096094092

09088086084082

08

(c) ROC Strategy 2 plaza scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC train data learnt offline

LawnPlazaIndoor

(d) ROC train offline

Figure 8 ROC curve of UMN dataset (a) ROC curve of different features 119865 results via Strategy 2 of the lawn sceneThe biggest AUC value is09605 (b) Strategy 2 results of the indoor scene The biggest AUC value is 08495 (c) Strategy 2 results of the plaza scene The biggest AUCvalue is 09746 (d)The ROC curve of best performance of the lawn indoor and plaza scene when the training samples are learnt offlineThebiggest AUC values of the lawn indoor and plaza are 09591 08649 and 09649

the dictionary and the classifier are not changed Otherwiseif the sample is classified as a normal one the sparse criterionintroduced in Section 3 is used to check the correlationbetween the earlier dictionary and this new datum It willbe included into the dictionary when it satisfied the updatecondition The dictionary will be updated through the wholetesting period The other two scenes the indoor and plazascene are handled by the same methods When 119865

5-17 times 17

feature is adopted the variance of the Gaussian kernel is120590 = 1 and the preset threshold of the criterion is 120583

0= 05

and the dictionary size of the lawn indoor and plaza scene

is increased from 100 to 106 102 and 102 respectively TheROC curve of detection results of these three scenes is shownin Figures 8(a) 8(b) and 8(c) Besides the merit of savingmemory of Strategy 1 Strategy 2 also has the advantage ofadaptation to the long duration sequence

The results performances of offline strategy Strategy 1and Strategy 2 are shown in Table 2 The performances ofthese two strategies results are similar to that of the resultswhen all training samples are learnt together When 119865

4(12 times

12) or1198655(17times17) are chosen as the features to formcovariance

matrix descriptor the results have the best performance

Journal of Electrical and Computer Engineering 11

These two features are more abundant to include movementand intensity information

The result performances of the covariancematrix descrip-tor based online one-class SVM method and the state-of-the-art methods are shown in Table 3 The covariance matrixbased online abnormal frame detection method obtainscompetitive performance In general our method is betterthan others except sparse reconstruction cost (SRC) [17]in lawn scene and indoor scene In that paper multiscaleHOF is taken as a feature and a testing sample is classi-fied by its sparse reconstructor cost through a weightedlinear reconstruction of the overcomplete normal basis setBut computation of the HOF might takes more time thancalculating covariance By adopting the integral image [15]the covariance matrix descriptor of the subimage can becomputed conveniently So the covariance descriptor can beappropriately used to analyze the partial movement

6 Conclusions

A method for the abnormal event detection of the frameis proposed The method consists of covariance matrixdescriptor encoding the movement features and the onlinenonlinear one-class SVM classification method We havedeveloped two nonlinear one-class SVM based abnormalevent detection techniques that update the normal modelsof the surveillance video data in an online framework Theproposed algorithm has been tested on a video datasetyielding successful results to detect abnormal events

Acknowledgments

This work is partially supported by China ScholarshipCouncil of Chinese Government SURECAP CPER Projectfunded by Region Champagne-Ardenne and CAPSECCRCA FEDER platform

References

[1] F Jiang J Yuan S A Tsaftaris and A K Katsaggelos ldquoAnoma-lous video event detection using spatiotemporal contextrdquo Com-puter Vision and Image Understanding vol 115 no 3 pp 323ndash333 2011

[2] C Piciarelli C Micheloni and G L Foresti ldquoTrajectory-basedanomalous event detectionrdquo IEEE Transactions on Circuits andSystems for Video Technology vol 18 no 11 pp 1544ndash1554 2008

[3] C Piciarelli and G L Foresti ldquoOn-line trajectory clustering foranomalous events detectionrdquo Pattern Recognition Letters vol27 no 15 pp 1835ndash1842 2006

[4] T Xiang and S Gong ldquoIncremental and adaptive abnormalbehaviour detectionrdquo Computer Vision and Image Understand-ing vol 111 no 1 pp 59ndash73 2008

[5] S Gong and T Xiang ldquoRecognition of group activities usingdynamic probabilistic networksrdquo in Proceedings of the 9th IEEEInternational Conference on Computer Vision (ICCV rsquo03) pp742ndash749 October 2003

[6] J KimandKGrauman ldquoObserve locally infer globally a space-time MRF for detecting abnormal activities with incrementalupdatesrdquo in Proceedings of the IEEE Conference on Computer

Vision and Pattern Recognition (CVPR rsquo09) pp 2921ndash2928Miami Fla USA June 2009

[7] A Adam E Rivlin I Shimshoni and D Reinitz ldquoRobustreal-time unusual event detection using multiple fixed-locationmonitorsrdquo IEEE Transactions on Pattern Analysis and MachineIntelligence vol 30 no 3 pp 555ndash560 2008

[8] O Boiman andM Irani ldquoDetecting irregularities in images andin videordquo International Journal of Computer Vision vol 74 no1 pp 17ndash31 2007

[9] X Zhu Z Liu and J Zhang ldquoHuman activity clustering foronline anomaly detectionrdquo Journal of Computers vol 6 no 6pp 1071ndash1079 2011

[10] X Zhu and Z Liu ldquoHuman behavior clustering for anomalydetectionrdquo Frontiers of Computer Science in China vol 5 no3 pp 279ndash289 2011

[11] Y Benezeth P M Jodoin and V Saligrama ldquoAbnormalitydetection using low-level co-occurring eventsrdquo Pattern Recog-nition Letters vol 32 no 3 pp 423ndash431 2011

[12] Y Benezeth P M Jodoin V Saligrama and C RosenbergerldquoAbnormal events detection based on spatio-temporal co-occurencesrdquo in Proceedings of the IEEE Conference on ComputerVision and Pattern Recognition (CVPR rsquo09) pp 2458ndash2465Miami Fla USA June 2009

[13] A Mittal A Monnet and N Paragios ldquoScene modeling andchange detection in dynamic scenes a subspace approachrdquoComputer Vision and Image Understanding vol 113 no 1 pp63ndash79 2009

[14] B K P Horn and B G Schunck ldquoDetermining optical flowrdquoArtificial Intelligence vol 17 no 1ndash3 pp 185ndash203 1981

[15] O Tuzel F Porikli and P Meer ldquoRegion covariance afast descriptor for detection and classificationrdquo in ComputerVisionmdashECCV 2006 vol 3952 of Lecture Notes in ComputerScience pp 589ndash600 Springer New York NY USA 2006

[16] RMehran A Oyama andM Shah ldquoAbnormal crowd behaviordetection using social force modelrdquo in Proceedings of the IEEEConference on Computer Vision and Pattern Recognition (CVPRrsquo09) pp 935ndash942 Miami Fla USA June 2009

[17] Y Cong J Yuan and J Liu ldquoSparse reconstruction cost forabnormal event detectionrdquo in Proceedings of the IEEE Confer-ence on Computer Vision and Pattern Recognition (CVPR rsquo11)pp 3449ndash3456 Providence RI USA June 2011

[18] Y Shi Y Gao and R Wang ldquoReal-time abnormal eventdetection in complicated scenesrdquo in Proceedings of the 20thInternational Conference on Pattern Recognition (ICPR rsquo10) pp3653ndash3656 Istanbul Turkey August 2010

[19] B Scholkopf and A J Smola Learning with Kernels SupportVector Machines Regularization Optimization and BeyondMIT Press Boston Mass USA 2002

[20] B Hall Lie Groups Lie Algebras And Representations AnElementary Introduction vol 222 Springer New York NYUSA 2003

[21] C Gao F Li and C Shen ldquoResearch on lie group kernellearning algorithmrdquo Journal of Frontiers of Compter Science andTechnology vol 6 pp 1026ndash1038 2012

[22] V N Vapnik and A Lerner ldquoPattern recognition using general-ized portrait methodrdquo Automation and Remote Control vol 24pp 774ndash780 1963

[23] V Vapnik The Nature of Statistical Learning Theory SpringerNew York NY USA 2000

[24] D TaxOne-class classification [PhD thesis] Delft University ofTechnology 2001

12 Journal of Electrical and Computer Engineering

[25] D M Tax and R P Duin ldquoData domain description usingsupport vectorsrdquo in Proceedings of the European Symposiumon Artificial Neural Networks Computational Intelligence andMachine Learning (ESANN rsquo99) vol 99 pp 251ndash256 1999

[26] B Scholkopf J C Platt J Shawe-Taylor A J Smola and RC Williamson ldquoEstimating the support of a high-dimensionaldistributionrdquo Neural Computation vol 13 no 7 pp 1443ndash14712001

[27] Z Noumir P Honeine and C Richard ldquoOnline one-classmachines based on the coherence criterionrdquo in Proceedings ofthe 20th European Signal Processing Conference (EUSIPCO rsquo12)pp 664ndash668 2012

[28] Z Noumir P Honeine and C Richard ldquoOne-class machinesbased on the coherence criterionrdquo in Proceedings of the IEEEStatistical Signal Processing Workshop (SSP rsquo12) pp 600ndash6032012

[29] P Honeine ldquoOnline kernel principal component analysis areducedorder modelrdquo IEEE Transactions on Pattern Analysisand Machine Intelligence vol 34 no 9 pp 1814ndash1826 2012

[30] C Richard J C M Bermudez and P Honeine ldquoOnlineprediction of time series data with kernelsrdquo IEEE Transactionson Signal Processing vol 57 no 3 pp 1058ndash1067 2009

[31] UMN ldquoUnusual crowd activity dataset of university of Min-nesotardquo 2006 httpmhacsumneduproj eventsshtml

[32] J A Hanley and B J McNeil ldquoThe meaning and use of thearea under a receiver operating characteristic (ROC) curverdquoRadiology vol 143 no 1 pp 29ndash36 1982

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Journal of Electrical and Computer Engineering 9

(a) Normal plaza scene (b) Abnormal plaza scene

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

ROC plaza offline

False positive

True

pos

itive

F1

F2

F3

F4

F5

(c) ROC train offline

1

08

06

04

02

00 02 04 06 08 1

098096094092

09088086084082

080 005 01 015 02

False positive

True

pos

itive

ROC plaza Strategy 1

F1

F2

F3

F4

F5

(d) ROC Strategy 1

Figure 7 Detection results of the plaza scene (a) The detection result of one normal frame (b) The detection result of one abnormal panicframe (c) ROC curve of different features 119865 of the plaza scene results via one-class SVM All the training samples are learned together offlineThe biggest AUC value is 09649 (d) ROC curve of different features 119865 results via Strategy 1 online one-class SVMThe biggest AUC value is09632

Online one-class SVM is appropriate to detect abnormalvisual events There are 1431 frames in the lawn scene and480 normal frames are used for training In the offline strat-egy all the 480 frames covariancematrices should be saved inthememory In Strategy 1 100 frames covariancematrices areconsidered as the dictionary firstlyWhen119865

5-17times17 feature is

adopted to construct the covariance descriptor the varianceof Gaussian kernel is 120590 = 1 the preset threshold of thecriterion is 120583

0= 05 the dictionary size increases from 100

to 101 and the maximum accuracy of the detection resultsis 9169 In the indoor scene there are 2975 normal framesand 1057 abnormal frames In the plaza scene there are 1831normal frames and 286 abnormal framesTheprocesses of theexperiments are similar to the ones of the lawn scene Whenfeature vector is 119865

5-17 times 17 120590 = 1 120583

0= 05 the dictionary

size of these two scenes remain 100The online strategy keepsthe memory size almost unchanged when the size of trainingdataset increases

52 Abnormal Visual Events Detection Strategy 2 The resultsof the abnormal event detection method via Strategy 2 ofUMNdataset are shown as follows In the experiment processof the lawn scene 100 normal samples from the trainingsamples are learnt firstly and then the other 380 trainingdata are learnt online one by one After these two trainingsteps we can obtain the basic dictionary from the trainingsamples and also the classifier In the following testing stepthe dictionary is updated if the sample satisfies the dictionaryupdate criterion When a new sample is coming it is firstlydetected by the previous classifier If it is classified as anomaly

10 Journal of Electrical and Computer Engineering

094092

09088086084082

08

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC lawn Strategy 2

F1

F2

F3

F4

F5

(a) ROC Strategy 2 lawn scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

085

08

075

07

065

0 01 02 03 04

ROC indoor Strategy 2

F1

F2

F3

F4

F5

(b) ROC Strategy 2 indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC plaza Strategy 2

F1

F2

F3

F4

F5

098096094092

09088086084082

08

(c) ROC Strategy 2 plaza scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC train data learnt offline

LawnPlazaIndoor

(d) ROC train offline

Figure 8 ROC curve of UMN dataset (a) ROC curve of different features 119865 results via Strategy 2 of the lawn sceneThe biggest AUC value is09605 (b) Strategy 2 results of the indoor scene The biggest AUC value is 08495 (c) Strategy 2 results of the plaza scene The biggest AUCvalue is 09746 (d)The ROC curve of best performance of the lawn indoor and plaza scene when the training samples are learnt offlineThebiggest AUC values of the lawn indoor and plaza are 09591 08649 and 09649

the dictionary and the classifier are not changed Otherwiseif the sample is classified as a normal one the sparse criterionintroduced in Section 3 is used to check the correlationbetween the earlier dictionary and this new datum It willbe included into the dictionary when it satisfied the updatecondition The dictionary will be updated through the wholetesting period The other two scenes the indoor and plazascene are handled by the same methods When 119865

5-17 times 17

feature is adopted the variance of the Gaussian kernel is120590 = 1 and the preset threshold of the criterion is 120583

0= 05

and the dictionary size of the lawn indoor and plaza scene

is increased from 100 to 106 102 and 102 respectively TheROC curve of detection results of these three scenes is shownin Figures 8(a) 8(b) and 8(c) Besides the merit of savingmemory of Strategy 1 Strategy 2 also has the advantage ofadaptation to the long duration sequence

The results performances of offline strategy Strategy 1and Strategy 2 are shown in Table 2 The performances ofthese two strategies results are similar to that of the resultswhen all training samples are learnt together When 119865

4(12 times

12) or1198655(17times17) are chosen as the features to formcovariance

matrix descriptor the results have the best performance

Journal of Electrical and Computer Engineering 11

These two features are more abundant to include movementand intensity information

The result performances of the covariancematrix descrip-tor based online one-class SVM method and the state-of-the-art methods are shown in Table 3 The covariance matrixbased online abnormal frame detection method obtainscompetitive performance In general our method is betterthan others except sparse reconstruction cost (SRC) [17]in lawn scene and indoor scene In that paper multiscaleHOF is taken as a feature and a testing sample is classi-fied by its sparse reconstructor cost through a weightedlinear reconstruction of the overcomplete normal basis setBut computation of the HOF might takes more time thancalculating covariance By adopting the integral image [15]the covariance matrix descriptor of the subimage can becomputed conveniently So the covariance descriptor can beappropriately used to analyze the partial movement

6 Conclusions

A method for the abnormal event detection of the frameis proposed The method consists of covariance matrixdescriptor encoding the movement features and the onlinenonlinear one-class SVM classification method We havedeveloped two nonlinear one-class SVM based abnormalevent detection techniques that update the normal modelsof the surveillance video data in an online framework Theproposed algorithm has been tested on a video datasetyielding successful results to detect abnormal events

Acknowledgments

This work is partially supported by China ScholarshipCouncil of Chinese Government SURECAP CPER Projectfunded by Region Champagne-Ardenne and CAPSECCRCA FEDER platform

References

[1] F Jiang J Yuan S A Tsaftaris and A K Katsaggelos ldquoAnoma-lous video event detection using spatiotemporal contextrdquo Com-puter Vision and Image Understanding vol 115 no 3 pp 323ndash333 2011

[2] C Piciarelli C Micheloni and G L Foresti ldquoTrajectory-basedanomalous event detectionrdquo IEEE Transactions on Circuits andSystems for Video Technology vol 18 no 11 pp 1544ndash1554 2008

[3] C Piciarelli and G L Foresti ldquoOn-line trajectory clustering foranomalous events detectionrdquo Pattern Recognition Letters vol27 no 15 pp 1835ndash1842 2006

[4] T Xiang and S Gong ldquoIncremental and adaptive abnormalbehaviour detectionrdquo Computer Vision and Image Understand-ing vol 111 no 1 pp 59ndash73 2008

[5] S Gong and T Xiang ldquoRecognition of group activities usingdynamic probabilistic networksrdquo in Proceedings of the 9th IEEEInternational Conference on Computer Vision (ICCV rsquo03) pp742ndash749 October 2003

[6] J KimandKGrauman ldquoObserve locally infer globally a space-time MRF for detecting abnormal activities with incrementalupdatesrdquo in Proceedings of the IEEE Conference on Computer

Vision and Pattern Recognition (CVPR rsquo09) pp 2921ndash2928Miami Fla USA June 2009

[7] A Adam E Rivlin I Shimshoni and D Reinitz ldquoRobustreal-time unusual event detection using multiple fixed-locationmonitorsrdquo IEEE Transactions on Pattern Analysis and MachineIntelligence vol 30 no 3 pp 555ndash560 2008

[8] O Boiman andM Irani ldquoDetecting irregularities in images andin videordquo International Journal of Computer Vision vol 74 no1 pp 17ndash31 2007

[9] X Zhu Z Liu and J Zhang ldquoHuman activity clustering foronline anomaly detectionrdquo Journal of Computers vol 6 no 6pp 1071ndash1079 2011

[10] X Zhu and Z Liu ldquoHuman behavior clustering for anomalydetectionrdquo Frontiers of Computer Science in China vol 5 no3 pp 279ndash289 2011

[11] Y Benezeth P M Jodoin and V Saligrama ldquoAbnormalitydetection using low-level co-occurring eventsrdquo Pattern Recog-nition Letters vol 32 no 3 pp 423ndash431 2011

[12] Y Benezeth P M Jodoin V Saligrama and C RosenbergerldquoAbnormal events detection based on spatio-temporal co-occurencesrdquo in Proceedings of the IEEE Conference on ComputerVision and Pattern Recognition (CVPR rsquo09) pp 2458ndash2465Miami Fla USA June 2009

[13] A Mittal A Monnet and N Paragios ldquoScene modeling andchange detection in dynamic scenes a subspace approachrdquoComputer Vision and Image Understanding vol 113 no 1 pp63ndash79 2009

[14] B K P Horn and B G Schunck ldquoDetermining optical flowrdquoArtificial Intelligence vol 17 no 1ndash3 pp 185ndash203 1981

[15] O Tuzel F Porikli and P Meer ldquoRegion covariance afast descriptor for detection and classificationrdquo in ComputerVisionmdashECCV 2006 vol 3952 of Lecture Notes in ComputerScience pp 589ndash600 Springer New York NY USA 2006

[16] RMehran A Oyama andM Shah ldquoAbnormal crowd behaviordetection using social force modelrdquo in Proceedings of the IEEEConference on Computer Vision and Pattern Recognition (CVPRrsquo09) pp 935ndash942 Miami Fla USA June 2009

[17] Y Cong J Yuan and J Liu ldquoSparse reconstruction cost forabnormal event detectionrdquo in Proceedings of the IEEE Confer-ence on Computer Vision and Pattern Recognition (CVPR rsquo11)pp 3449ndash3456 Providence RI USA June 2011

[18] Y Shi Y Gao and R Wang ldquoReal-time abnormal eventdetection in complicated scenesrdquo in Proceedings of the 20thInternational Conference on Pattern Recognition (ICPR rsquo10) pp3653ndash3656 Istanbul Turkey August 2010

[19] B Scholkopf and A J Smola Learning with Kernels SupportVector Machines Regularization Optimization and BeyondMIT Press Boston Mass USA 2002

[20] B Hall Lie Groups Lie Algebras And Representations AnElementary Introduction vol 222 Springer New York NYUSA 2003

[21] C Gao F Li and C Shen ldquoResearch on lie group kernellearning algorithmrdquo Journal of Frontiers of Compter Science andTechnology vol 6 pp 1026ndash1038 2012

[22] V N Vapnik and A Lerner ldquoPattern recognition using general-ized portrait methodrdquo Automation and Remote Control vol 24pp 774ndash780 1963

[23] V Vapnik The Nature of Statistical Learning Theory SpringerNew York NY USA 2000

[24] D TaxOne-class classification [PhD thesis] Delft University ofTechnology 2001

12 Journal of Electrical and Computer Engineering

[25] D M Tax and R P Duin ldquoData domain description usingsupport vectorsrdquo in Proceedings of the European Symposiumon Artificial Neural Networks Computational Intelligence andMachine Learning (ESANN rsquo99) vol 99 pp 251ndash256 1999

[26] B Scholkopf J C Platt J Shawe-Taylor A J Smola and RC Williamson ldquoEstimating the support of a high-dimensionaldistributionrdquo Neural Computation vol 13 no 7 pp 1443ndash14712001

[27] Z Noumir P Honeine and C Richard ldquoOnline one-classmachines based on the coherence criterionrdquo in Proceedings ofthe 20th European Signal Processing Conference (EUSIPCO rsquo12)pp 664ndash668 2012

[28] Z Noumir P Honeine and C Richard ldquoOne-class machinesbased on the coherence criterionrdquo in Proceedings of the IEEEStatistical Signal Processing Workshop (SSP rsquo12) pp 600ndash6032012

[29] P Honeine ldquoOnline kernel principal component analysis areducedorder modelrdquo IEEE Transactions on Pattern Analysisand Machine Intelligence vol 34 no 9 pp 1814ndash1826 2012

[30] C Richard J C M Bermudez and P Honeine ldquoOnlineprediction of time series data with kernelsrdquo IEEE Transactionson Signal Processing vol 57 no 3 pp 1058ndash1067 2009

[31] UMN ldquoUnusual crowd activity dataset of university of Min-nesotardquo 2006 httpmhacsumneduproj eventsshtml

[32] J A Hanley and B J McNeil ldquoThe meaning and use of thearea under a receiver operating characteristic (ROC) curverdquoRadiology vol 143 no 1 pp 29ndash36 1982

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

10 Journal of Electrical and Computer Engineering

094092

09088086084082

08

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC lawn Strategy 2

F1

F2

F3

F4

F5

(a) ROC Strategy 2 lawn scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

085

08

075

07

065

0 01 02 03 04

ROC indoor Strategy 2

F1

F2

F3

F4

F5

(b) ROC Strategy 2 indoor scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

0 005 01 015 02

ROC plaza Strategy 2

F1

F2

F3

F4

F5

098096094092

09088086084082

08

(c) ROC Strategy 2 plaza scene

1

08

06

04

02

00 02 04 06 08 1

False positive

True

pos

itive

ROC train data learnt offline

LawnPlazaIndoor

(d) ROC train offline

Figure 8 ROC curve of UMN dataset (a) ROC curve of different features 119865 results via Strategy 2 of the lawn sceneThe biggest AUC value is09605 (b) Strategy 2 results of the indoor scene The biggest AUC value is 08495 (c) Strategy 2 results of the plaza scene The biggest AUCvalue is 09746 (d)The ROC curve of best performance of the lawn indoor and plaza scene when the training samples are learnt offlineThebiggest AUC values of the lawn indoor and plaza are 09591 08649 and 09649

the dictionary and the classifier are not changed Otherwiseif the sample is classified as a normal one the sparse criterionintroduced in Section 3 is used to check the correlationbetween the earlier dictionary and this new datum It willbe included into the dictionary when it satisfied the updatecondition The dictionary will be updated through the wholetesting period The other two scenes the indoor and plazascene are handled by the same methods When 119865

5-17 times 17

feature is adopted the variance of the Gaussian kernel is120590 = 1 and the preset threshold of the criterion is 120583

0= 05

and the dictionary size of the lawn indoor and plaza scene

is increased from 100 to 106 102 and 102 respectively TheROC curve of detection results of these three scenes is shownin Figures 8(a) 8(b) and 8(c) Besides the merit of savingmemory of Strategy 1 Strategy 2 also has the advantage ofadaptation to the long duration sequence

The results performances of offline strategy Strategy 1and Strategy 2 are shown in Table 2 The performances ofthese two strategies results are similar to that of the resultswhen all training samples are learnt together When 119865

4(12 times

12) or1198655(17times17) are chosen as the features to formcovariance

matrix descriptor the results have the best performance

Journal of Electrical and Computer Engineering 11

These two features are more abundant to include movementand intensity information

The result performances of the covariancematrix descrip-tor based online one-class SVM method and the state-of-the-art methods are shown in Table 3 The covariance matrixbased online abnormal frame detection method obtainscompetitive performance In general our method is betterthan others except sparse reconstruction cost (SRC) [17]in lawn scene and indoor scene In that paper multiscaleHOF is taken as a feature and a testing sample is classi-fied by its sparse reconstructor cost through a weightedlinear reconstruction of the overcomplete normal basis setBut computation of the HOF might takes more time thancalculating covariance By adopting the integral image [15]the covariance matrix descriptor of the subimage can becomputed conveniently So the covariance descriptor can beappropriately used to analyze the partial movement

6 Conclusions

A method for the abnormal event detection of the frameis proposed The method consists of covariance matrixdescriptor encoding the movement features and the onlinenonlinear one-class SVM classification method We havedeveloped two nonlinear one-class SVM based abnormalevent detection techniques that update the normal modelsof the surveillance video data in an online framework Theproposed algorithm has been tested on a video datasetyielding successful results to detect abnormal events

Acknowledgments

This work is partially supported by China ScholarshipCouncil of Chinese Government SURECAP CPER Projectfunded by Region Champagne-Ardenne and CAPSECCRCA FEDER platform

References

[1] F Jiang J Yuan S A Tsaftaris and A K Katsaggelos ldquoAnoma-lous video event detection using spatiotemporal contextrdquo Com-puter Vision and Image Understanding vol 115 no 3 pp 323ndash333 2011

[2] C Piciarelli C Micheloni and G L Foresti ldquoTrajectory-basedanomalous event detectionrdquo IEEE Transactions on Circuits andSystems for Video Technology vol 18 no 11 pp 1544ndash1554 2008

[3] C Piciarelli and G L Foresti ldquoOn-line trajectory clustering foranomalous events detectionrdquo Pattern Recognition Letters vol27 no 15 pp 1835ndash1842 2006

[4] T Xiang and S Gong ldquoIncremental and adaptive abnormalbehaviour detectionrdquo Computer Vision and Image Understand-ing vol 111 no 1 pp 59ndash73 2008

[5] S Gong and T Xiang ldquoRecognition of group activities usingdynamic probabilistic networksrdquo in Proceedings of the 9th IEEEInternational Conference on Computer Vision (ICCV rsquo03) pp742ndash749 October 2003

[6] J KimandKGrauman ldquoObserve locally infer globally a space-time MRF for detecting abnormal activities with incrementalupdatesrdquo in Proceedings of the IEEE Conference on Computer

Vision and Pattern Recognition (CVPR rsquo09) pp 2921ndash2928Miami Fla USA June 2009

[7] A Adam E Rivlin I Shimshoni and D Reinitz ldquoRobustreal-time unusual event detection using multiple fixed-locationmonitorsrdquo IEEE Transactions on Pattern Analysis and MachineIntelligence vol 30 no 3 pp 555ndash560 2008

[8] O Boiman andM Irani ldquoDetecting irregularities in images andin videordquo International Journal of Computer Vision vol 74 no1 pp 17ndash31 2007

[9] X Zhu Z Liu and J Zhang ldquoHuman activity clustering foronline anomaly detectionrdquo Journal of Computers vol 6 no 6pp 1071ndash1079 2011

[10] X Zhu and Z Liu ldquoHuman behavior clustering for anomalydetectionrdquo Frontiers of Computer Science in China vol 5 no3 pp 279ndash289 2011

[11] Y Benezeth P M Jodoin and V Saligrama ldquoAbnormalitydetection using low-level co-occurring eventsrdquo Pattern Recog-nition Letters vol 32 no 3 pp 423ndash431 2011

[12] Y Benezeth P M Jodoin V Saligrama and C RosenbergerldquoAbnormal events detection based on spatio-temporal co-occurencesrdquo in Proceedings of the IEEE Conference on ComputerVision and Pattern Recognition (CVPR rsquo09) pp 2458ndash2465Miami Fla USA June 2009

[13] A Mittal A Monnet and N Paragios ldquoScene modeling andchange detection in dynamic scenes a subspace approachrdquoComputer Vision and Image Understanding vol 113 no 1 pp63ndash79 2009

[14] B K P Horn and B G Schunck ldquoDetermining optical flowrdquoArtificial Intelligence vol 17 no 1ndash3 pp 185ndash203 1981

[15] O Tuzel F Porikli and P Meer ldquoRegion covariance afast descriptor for detection and classificationrdquo in ComputerVisionmdashECCV 2006 vol 3952 of Lecture Notes in ComputerScience pp 589ndash600 Springer New York NY USA 2006

[16] RMehran A Oyama andM Shah ldquoAbnormal crowd behaviordetection using social force modelrdquo in Proceedings of the IEEEConference on Computer Vision and Pattern Recognition (CVPRrsquo09) pp 935ndash942 Miami Fla USA June 2009

[17] Y Cong J Yuan and J Liu ldquoSparse reconstruction cost forabnormal event detectionrdquo in Proceedings of the IEEE Confer-ence on Computer Vision and Pattern Recognition (CVPR rsquo11)pp 3449ndash3456 Providence RI USA June 2011

[18] Y Shi Y Gao and R Wang ldquoReal-time abnormal eventdetection in complicated scenesrdquo in Proceedings of the 20thInternational Conference on Pattern Recognition (ICPR rsquo10) pp3653ndash3656 Istanbul Turkey August 2010

[19] B Scholkopf and A J Smola Learning with Kernels SupportVector Machines Regularization Optimization and BeyondMIT Press Boston Mass USA 2002

[20] B Hall Lie Groups Lie Algebras And Representations AnElementary Introduction vol 222 Springer New York NYUSA 2003

[21] C Gao F Li and C Shen ldquoResearch on lie group kernellearning algorithmrdquo Journal of Frontiers of Compter Science andTechnology vol 6 pp 1026ndash1038 2012

[22] V N Vapnik and A Lerner ldquoPattern recognition using general-ized portrait methodrdquo Automation and Remote Control vol 24pp 774ndash780 1963

[23] V Vapnik The Nature of Statistical Learning Theory SpringerNew York NY USA 2000

[24] D TaxOne-class classification [PhD thesis] Delft University ofTechnology 2001

12 Journal of Electrical and Computer Engineering

[25] D M Tax and R P Duin ldquoData domain description usingsupport vectorsrdquo in Proceedings of the European Symposiumon Artificial Neural Networks Computational Intelligence andMachine Learning (ESANN rsquo99) vol 99 pp 251ndash256 1999

[26] B Scholkopf J C Platt J Shawe-Taylor A J Smola and RC Williamson ldquoEstimating the support of a high-dimensionaldistributionrdquo Neural Computation vol 13 no 7 pp 1443ndash14712001

[27] Z Noumir P Honeine and C Richard ldquoOnline one-classmachines based on the coherence criterionrdquo in Proceedings ofthe 20th European Signal Processing Conference (EUSIPCO rsquo12)pp 664ndash668 2012

[28] Z Noumir P Honeine and C Richard ldquoOne-class machinesbased on the coherence criterionrdquo in Proceedings of the IEEEStatistical Signal Processing Workshop (SSP rsquo12) pp 600ndash6032012

[29] P Honeine ldquoOnline kernel principal component analysis areducedorder modelrdquo IEEE Transactions on Pattern Analysisand Machine Intelligence vol 34 no 9 pp 1814ndash1826 2012

[30] C Richard J C M Bermudez and P Honeine ldquoOnlineprediction of time series data with kernelsrdquo IEEE Transactionson Signal Processing vol 57 no 3 pp 1058ndash1067 2009

[31] UMN ldquoUnusual crowd activity dataset of university of Min-nesotardquo 2006 httpmhacsumneduproj eventsshtml

[32] J A Hanley and B J McNeil ldquoThe meaning and use of thearea under a receiver operating characteristic (ROC) curverdquoRadiology vol 143 no 1 pp 29ndash36 1982

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Journal of Electrical and Computer Engineering 11

These two features are more abundant to include movementand intensity information

The result performances of the covariancematrix descrip-tor based online one-class SVM method and the state-of-the-art methods are shown in Table 3 The covariance matrixbased online abnormal frame detection method obtainscompetitive performance In general our method is betterthan others except sparse reconstruction cost (SRC) [17]in lawn scene and indoor scene In that paper multiscaleHOF is taken as a feature and a testing sample is classi-fied by its sparse reconstructor cost through a weightedlinear reconstruction of the overcomplete normal basis setBut computation of the HOF might takes more time thancalculating covariance By adopting the integral image [15]the covariance matrix descriptor of the subimage can becomputed conveniently So the covariance descriptor can beappropriately used to analyze the partial movement

6 Conclusions

A method for the abnormal event detection of the frameis proposed The method consists of covariance matrixdescriptor encoding the movement features and the onlinenonlinear one-class SVM classification method We havedeveloped two nonlinear one-class SVM based abnormalevent detection techniques that update the normal modelsof the surveillance video data in an online framework Theproposed algorithm has been tested on a video datasetyielding successful results to detect abnormal events

Acknowledgments

This work is partially supported by China ScholarshipCouncil of Chinese Government SURECAP CPER Projectfunded by Region Champagne-Ardenne and CAPSECCRCA FEDER platform

References

[1] F Jiang J Yuan S A Tsaftaris and A K Katsaggelos ldquoAnoma-lous video event detection using spatiotemporal contextrdquo Com-puter Vision and Image Understanding vol 115 no 3 pp 323ndash333 2011

[2] C Piciarelli C Micheloni and G L Foresti ldquoTrajectory-basedanomalous event detectionrdquo IEEE Transactions on Circuits andSystems for Video Technology vol 18 no 11 pp 1544ndash1554 2008

[3] C Piciarelli and G L Foresti ldquoOn-line trajectory clustering foranomalous events detectionrdquo Pattern Recognition Letters vol27 no 15 pp 1835ndash1842 2006

[4] T Xiang and S Gong ldquoIncremental and adaptive abnormalbehaviour detectionrdquo Computer Vision and Image Understand-ing vol 111 no 1 pp 59ndash73 2008

[5] S Gong and T Xiang ldquoRecognition of group activities usingdynamic probabilistic networksrdquo in Proceedings of the 9th IEEEInternational Conference on Computer Vision (ICCV rsquo03) pp742ndash749 October 2003

[6] J KimandKGrauman ldquoObserve locally infer globally a space-time MRF for detecting abnormal activities with incrementalupdatesrdquo in Proceedings of the IEEE Conference on Computer

Vision and Pattern Recognition (CVPR rsquo09) pp 2921ndash2928Miami Fla USA June 2009

[7] A Adam E Rivlin I Shimshoni and D Reinitz ldquoRobustreal-time unusual event detection using multiple fixed-locationmonitorsrdquo IEEE Transactions on Pattern Analysis and MachineIntelligence vol 30 no 3 pp 555ndash560 2008

[8] O Boiman andM Irani ldquoDetecting irregularities in images andin videordquo International Journal of Computer Vision vol 74 no1 pp 17ndash31 2007

[9] X Zhu Z Liu and J Zhang ldquoHuman activity clustering foronline anomaly detectionrdquo Journal of Computers vol 6 no 6pp 1071ndash1079 2011

[10] X Zhu and Z Liu ldquoHuman behavior clustering for anomalydetectionrdquo Frontiers of Computer Science in China vol 5 no3 pp 279ndash289 2011

[11] Y Benezeth P M Jodoin and V Saligrama ldquoAbnormalitydetection using low-level co-occurring eventsrdquo Pattern Recog-nition Letters vol 32 no 3 pp 423ndash431 2011

[12] Y Benezeth P M Jodoin V Saligrama and C RosenbergerldquoAbnormal events detection based on spatio-temporal co-occurencesrdquo in Proceedings of the IEEE Conference on ComputerVision and Pattern Recognition (CVPR rsquo09) pp 2458ndash2465Miami Fla USA June 2009

[13] A Mittal A Monnet and N Paragios ldquoScene modeling andchange detection in dynamic scenes a subspace approachrdquoComputer Vision and Image Understanding vol 113 no 1 pp63ndash79 2009

[14] B K P Horn and B G Schunck ldquoDetermining optical flowrdquoArtificial Intelligence vol 17 no 1ndash3 pp 185ndash203 1981

[15] O Tuzel F Porikli and P Meer ldquoRegion covariance afast descriptor for detection and classificationrdquo in ComputerVisionmdashECCV 2006 vol 3952 of Lecture Notes in ComputerScience pp 589ndash600 Springer New York NY USA 2006

[16] RMehran A Oyama andM Shah ldquoAbnormal crowd behaviordetection using social force modelrdquo in Proceedings of the IEEEConference on Computer Vision and Pattern Recognition (CVPRrsquo09) pp 935ndash942 Miami Fla USA June 2009

[17] Y Cong J Yuan and J Liu ldquoSparse reconstruction cost forabnormal event detectionrdquo in Proceedings of the IEEE Confer-ence on Computer Vision and Pattern Recognition (CVPR rsquo11)pp 3449ndash3456 Providence RI USA June 2011

[18] Y Shi Y Gao and R Wang ldquoReal-time abnormal eventdetection in complicated scenesrdquo in Proceedings of the 20thInternational Conference on Pattern Recognition (ICPR rsquo10) pp3653ndash3656 Istanbul Turkey August 2010

[19] B Scholkopf and A J Smola Learning with Kernels SupportVector Machines Regularization Optimization and BeyondMIT Press Boston Mass USA 2002

[20] B Hall Lie Groups Lie Algebras And Representations AnElementary Introduction vol 222 Springer New York NYUSA 2003

[21] C Gao F Li and C Shen ldquoResearch on lie group kernellearning algorithmrdquo Journal of Frontiers of Compter Science andTechnology vol 6 pp 1026ndash1038 2012

[22] V N Vapnik and A Lerner ldquoPattern recognition using general-ized portrait methodrdquo Automation and Remote Control vol 24pp 774ndash780 1963

[23] V Vapnik The Nature of Statistical Learning Theory SpringerNew York NY USA 2000

[24] D TaxOne-class classification [PhD thesis] Delft University ofTechnology 2001

12 Journal of Electrical and Computer Engineering

[25] D M Tax and R P Duin ldquoData domain description usingsupport vectorsrdquo in Proceedings of the European Symposiumon Artificial Neural Networks Computational Intelligence andMachine Learning (ESANN rsquo99) vol 99 pp 251ndash256 1999

[26] B Scholkopf J C Platt J Shawe-Taylor A J Smola and RC Williamson ldquoEstimating the support of a high-dimensionaldistributionrdquo Neural Computation vol 13 no 7 pp 1443ndash14712001

[27] Z Noumir P Honeine and C Richard ldquoOnline one-classmachines based on the coherence criterionrdquo in Proceedings ofthe 20th European Signal Processing Conference (EUSIPCO rsquo12)pp 664ndash668 2012

[28] Z Noumir P Honeine and C Richard ldquoOne-class machinesbased on the coherence criterionrdquo in Proceedings of the IEEEStatistical Signal Processing Workshop (SSP rsquo12) pp 600ndash6032012

[29] P Honeine ldquoOnline kernel principal component analysis areducedorder modelrdquo IEEE Transactions on Pattern Analysisand Machine Intelligence vol 34 no 9 pp 1814ndash1826 2012

[30] C Richard J C M Bermudez and P Honeine ldquoOnlineprediction of time series data with kernelsrdquo IEEE Transactionson Signal Processing vol 57 no 3 pp 1058ndash1067 2009

[31] UMN ldquoUnusual crowd activity dataset of university of Min-nesotardquo 2006 httpmhacsumneduproj eventsshtml

[32] J A Hanley and B J McNeil ldquoThe meaning and use of thearea under a receiver operating characteristic (ROC) curverdquoRadiology vol 143 no 1 pp 29ndash36 1982

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

12 Journal of Electrical and Computer Engineering

[25] D M Tax and R P Duin ldquoData domain description usingsupport vectorsrdquo in Proceedings of the European Symposiumon Artificial Neural Networks Computational Intelligence andMachine Learning (ESANN rsquo99) vol 99 pp 251ndash256 1999

[26] B Scholkopf J C Platt J Shawe-Taylor A J Smola and RC Williamson ldquoEstimating the support of a high-dimensionaldistributionrdquo Neural Computation vol 13 no 7 pp 1443ndash14712001

[27] Z Noumir P Honeine and C Richard ldquoOnline one-classmachines based on the coherence criterionrdquo in Proceedings ofthe 20th European Signal Processing Conference (EUSIPCO rsquo12)pp 664ndash668 2012

[28] Z Noumir P Honeine and C Richard ldquoOne-class machinesbased on the coherence criterionrdquo in Proceedings of the IEEEStatistical Signal Processing Workshop (SSP rsquo12) pp 600ndash6032012

[29] P Honeine ldquoOnline kernel principal component analysis areducedorder modelrdquo IEEE Transactions on Pattern Analysisand Machine Intelligence vol 34 no 9 pp 1814ndash1826 2012

[30] C Richard J C M Bermudez and P Honeine ldquoOnlineprediction of time series data with kernelsrdquo IEEE Transactionson Signal Processing vol 57 no 3 pp 1058ndash1067 2009

[31] UMN ldquoUnusual crowd activity dataset of university of Min-nesotardquo 2006 httpmhacsumneduproj eventsshtml

[32] J A Hanley and B J McNeil ldquoThe meaning and use of thearea under a receiver operating characteristic (ROC) curverdquoRadiology vol 143 no 1 pp 29ndash36 1982

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of