Research Article Spectral Pattern Classification in …downloads.hindawi.com › journals › tswj › 2014 › 539029.pdfCac¸apava do Sul, approximately km from Porto Alegre with

Research ArticleSpectral Pattern Classification in Lidar Data for RockIdentification in Outcrops

Leonardo Campos Inocencio1 Mauricio Roberto Veronez12

Francisco Manoel Wohnrath Tognoli1 Marcelo Kehl de Souza1

Reginaldo Macedocircnio da Silva1 Luiz Gonzaga Jr34 and Ceacutesar Leonardo Blum Silveira4

1 VIZLab Advanced Visualization Laboratory UNISINOS 93022-000 Sao Leopoldo RS Brazil2 PPGEO Graduate Program on Geology UNISINOS 93022-000 Sao Leopoldo RS Brazil3 PIPCA Applied Computer Science Graduate Program UNISINOS 93022-000 Sao Leopoldo RS Brazil4V3D Studios amp Ficta Mobile Technologies 93022-000 Sao Leopoldo RS Brazil

Correspondence should be addressed to Mauricio Roberto Veronez veronezunisinosbr

Received 8 August 2013 Accepted 21 November 2013 Published 18 February 2014

Academic Editors P Fonseca K Nemeth C M Petrone and L Tosi

Copyright copy 2014 Leonardo Campos Inocencio et al This is an open access article distributed under the Creative CommonsAttribution License which permits unrestricted use distribution and reproduction in any medium provided the original work isproperly cited

The present study aimed to develop and implement a method for detection and classification of spectral signatures in point cloudsobtained from terrestrial laser scanner in order to identify the presence of different rocks in outcrops and to generate a digitaloutcrop model To achieve this objective a software based on cluster analysis was created named K-Clouds This software wasdeveloped through a partnership between UNISINOS and the company V3DThis tool was designed to begin with an analysis andinterpretation of a histogram from a point cloud of the outcrop and subsequently indication of a number of classes provided bythe user to process the intensity return values This classified information can then be interpreted by geologists to provide a betterunderstanding and identification from the existing rocks in the outcrop Beyond the detection of different rocks this work was ableto detect small changes in the physical-chemical characteristics of the rocks as they were caused by weathering or compositionalchanges

1 Introduction

Geology like other sciences enjoys technological advances toincreasingly develop theirmethods and enhance the expertiseof their field of study New equipment and methods arein constant development to support these applications andfrom all of this equipment and systems developed in recentyears the laser scanning and profiling has been consolidatedas one of the most effective technology for geospatial dataacquisition

The automated data collection has expanded rapidly inrecent years in line with the technological advances madein the areas of surveying and mapping [1] The use of suchequipment for geological studies was a natural and logicalstep considering that the collection and use of data fromoutcrops are themain source for the works of surface geology

The digital representation in geological studies evolvedfrom LANDSAT satellite imagery digital terrain models inconjunction with image processing (ie [2]) and aerial pho-tos interpretation techniques combinedwith theGlobalNavi-gation Satellite System (GNSS) (ie [3]) Over the past decadethe use of digital mapping technologies have increased inparticular with the use of terrestrial laser scanner surveyingand other systems integrated with satellite navigation andgeographic information [4ndash6] which are replaced with manyadvantages of photographic mosaics routinely used in theinterpretation of large outcrops

The laser scanning and profiling systems also known asterrestrial laser scanner (TLS) have some characteristics thatapply significantly for geological purposes such as the fastacquisition of data of a particular outcrop in distances thatcan reach 2 km the high definition of the spatial resolution

Hindawi Publishing Corporatione Scientific World JournalVolume 2014 Article ID 539029 10 pageshttpdxdoiorg1011552014539029

2 The Scientific World Journal

that is the spacing between the points collected at a certaindistance and high degree of accuracy of the survey

The use of TLS in outcrop studies is expanding due tothe ease of acquisition of georeferenced data in an accuratefast and automated way Its use for this purpose beganabout a decade ago [7] but only in recent years the numberof scientific papers has increased significantly Topics ofinterest are varied with emphasis on the methodologicalapproaches [1 8ndash11] reservoirs analogues studies [12ndash19] offractured rocks [20ndash23] slope stability [24] erosion rates[25] synthetic seismic models [26] geological heritage [27]determining the direction of basaltic lava flow [28] geologyeducation [29 30] and outcrop interpretation [31]

The equipment available on the market is able to conductthe survey and collection of thousands (time of flight) upto millions (phase) of points per second and some have theability to record the intensity of the pulse return for eachposition of the laser XYZ [8] The intensity of the returnpulse from the laser and the equipment range are directlyrelated to factors that affect theworking of the systemAmongthese factors we can mention the power of the emitted pulseand the reflectivity of the targets

The relation between the reflectivity of the targets and theintensity of the laser beam emitted is directly proportional tothe range of the Laser Scanner If a target has 10 reflectivitythe equipment should be closer than a target that has 90reflectivity [32]

As other active remote sensing systems the TLS has onlyone wavelength as the intensity of the return pulse is basedon the interaction of the laserrsquos specific wavelength with thetargetThis interactionmay be analyzed using interferometrytechniques which are the study of phenomena caused by theinterfering within the light waves which in this case are theinteraction between the laser wavelength and the physical-chemical characteristics of the targets [33]

As we use TLS to survey an outcrop we not only collecttheir spatial characteristics such as features and volumesbut we also obtain additional data that may allow a greaterunderstanding of the outcrop

In most cases each TLS equipment generates its own fileformat for archiving of the collected data However these fileshave similarities concerning the information stored Thesefiles record the spatial position of the points (X Y and Z) theintensity of the laser pulse that returns from the target (i) andthe RGB color pattern obtained from the pictures recordedby the internal camera The output is a txt file that providesthe sequence X Y Z I R G B

The spatial information is commonly used for volumecalculations and identification of geometric features and theRGB data assist in identifying visual features and provide abetter understanding of the objects surveyed with the TLSbut when it comes to the intensity values of the returnlaser pulse there is a lack of specific applications designedto transform this data in information which can detectvariations in the physical-chemical characteristics of thetargets

Thus the main objective of this study was to developa method for classification of spectral patterns in LIDAR

(light detect and range) data for the identification of rocks inoutcrops with distinct lithology

The validation of this method occurred through the useof a cluster analysis algorithm known as k-meansThe choiceof this algorithm was given by the obtained separation ofthe spectral responses of rocks with different characteristicsboth chemical andor physical This allowed us to classifythe patterns of spectral response and thus identify rocks andfeatures present in the outcrops This classifier tool will beincorporated in the software Mountain View (partnershipV3DUNISINOS) which allows the user view edits andprocesses data from TLS and the main application is focusedon Digital Outcrop Modeling (DOM)

2 Terrestrial Laser Scanner

The Laser Scanning and profiling systems in this caseterrestrial basically consist of a measuring instrument basedon laser that can measure vertical angles horizontal anglesand distances with a high standard of accuracy and speed bymeans of a mobile mirrors or prisms system that allow themapping of topographical features on targets

These systems have different methods for measure dis-tances and the time of flight (TOF) (Figure 1) and phase shift(PS) are the most widespread In the Time of Flight methodthe instrument sends a high-intensity laser pulse toward thetarget and measure the time between sending and receivingthe pulse returns and divides the value by 2 to calculate thedistance

In the phase shift method the instrument emanates alaser pulse fitted within a harmonic wave and the distance iscalculated by comparing the difference of the transmitted andreceived waves [35] (Figure 2)

Both methods of data acquisition have strengths andweaknesses due to the method applied The equipment thatuses the phase shift technology tends to collect more pointsper second at the same time interval millions of points persecond while the time of flight collects thousands of pointsper second However the phase shift scanners have a reducedrange that varies between 80 to 120 meters against up totwo kilometers in some time of flight equipment Anotherelement to consider is that the operation of phase shiftequipment in conditions of high insolation can reduce therange of the equipment

All the elements quoted above must be taken into con-sideration when selecting the most suitable technology toperform the work in which case Digital Outcrop Modeling

3 Classification

The term classification is defined as an actual or idealarrangement relating what is similar and separating what isdifferent having as main objectives to maintain and buildknowledge and analyze and relate the structure of certainphenomena [36] If we apply this definition to geologyspecifically in outcrops classification it becomes clear thatthe first and main elements to be classified and identified arethe rocks on site This information is intended to expand the

The Scientific World Journal 3

Emittedpulse

Primaryreturn pulse

Secondaryreturn pulse

Time

Inte

nsity

Figure 1 Time of flight measurement system operation principlewhere the distance is calculated dividing by 2 the time elapsedbetween the emitted pulse and the primary or secondary returnpulse Modified from San Jose Alonso et al [34]

knowledge of the geologists regarding the elements presentin the outcrop allowing a greater understanding of existingprocesses and materials

In digital image processing different classifiers are wellestablished as they are applied to specific purposes or appliedto general purpose or even allowing their use for applicationsin a classification tool that is both specific and generic

Works such as those developed by Brodu and Lague[37] and Franceschi et al [38] are intended to implementdifferent methods of classification in TLS data for extractinginformationThese classification tools allow a fast processingof database containing thousand to million points acquiredby TLS

In the various methods for data classification dataclustering also called cluster analysis segmentation analy-sis taxonomic analysis or unsupervised classification is amethod of creating groups of objects that at first performs thegrouping of all objects that have a high degree of similarityalways ensuring that very different objects do not belong tothe same group (cluster)

The data clustering techniques are often confused withclassification in which objects are assigned to predefinedclasses but in data clustering the classes must also be defined[39] Compared with other types of sorting algorithmsclusters based algorithms are very efficient for classificationof large databases and multidimensional data [39]

The classifier 119896-means first described by MacQueenin 1967 is a cluster classifier that performs a process ofpartitioning of a population ldquo119873rdquo into ldquo119896rdquo classes Thesepartitions represent satisfactorily the internal variability thatoccurred within each class Besides the described abovethe 119896-means classifier is easily programmable and computa-tionally economical being able to process large volumes ofdata in applications such as grouping by similarity predict-ing nonlinear approximation of multivariate distributionsand nonparametric tests among others [40] These featuresbecome suitable for use in applications related to geologywhere the study areas can be located inside a few metersor along several kilometers This scale variability togetherwith the type of data separation carried out by the classifier119896-means which is capable of displaying discrete patterns ofvariation allows a proper understanding of the data

x(t)

A

T

t

A(x t) = A120579 cos(wt minus kx + 120593)

Figure 2 Phase shift measurement system operation principlewhere the distance is calculated comparing the phase differencebetween the emitted pulse and the return pulse Modified from SanJose Alonso et al [34]

4 Study Areas

For the present work two study areas were selected in thecities of Mariana Pimentel and Cacapava do Sul both in RioGrande do Sul state (Figure 3) Beyond the study areas initialtests were also carried out on rock samples to define the workmethod to be used in the Laboratory of Remote Sensing andDigital Cartography (LASERCA) in Unisinos

The study area A is located in themunicipality ofMarianaPimentel approximately 80 km fromPortoAlegrewith accessfrom the highway BR-116 and after municipal and privateroads In this place the outcrop called ldquoMorro do Papaleordquo islocated at coordinates Latitude 30∘181015840295610158401015840 S and Longitude51∘381015840352610158401015840W Gr (SIRGAS 2000) This outcrop in the RioBonito Formation of Permian age was chosen by its distinctsedimentary rocks succession as diamictite carbonaceouspelite and sandstone respectively from bottom to top In thewest face of the outcrop an area of approximately 36 squaremeters (6m times 6m) was selected for testing the developedclassifier (Figure 4)

The study area B is located in the municipality ofCacapava do Sul approximately 330 km from Porto Alegrewith access by highway BR-290 BR-392 and RS-625 In thisplace the outcrop called ldquoPedra Pintadardquo is located in coordi-nates Latitude 30∘531015840562210158401015840 S and Longitude 53∘211015840371710158401015840WGr (SIRGAS 2000)This outcrop belongs to the Pedra PintadaAlloformation and consists of fine to medium sandstonewell sorted composed of well-rounded grains with highsphericity and structured in sets of cross-stratification up to15m thick [41] In this outcrop an area with 165 meters inlength was selected from the main face of the outcrop in thesouth side (Figure 5)

5 Materials and Methods

The operation of TLS equipment can be described as a simpleactivity but presents several details and requires specificknowledge of the purpose of use to allow a total capacityutilization of the equipment The planning of basic featuressuch as range both minimum and maximum expectedspectral behavior of the targets of interest and optimalresolution to be used is indispensable for obtaining three-dimensional outcrop data mainly data capable of being usedfor spectral classification


27∘S

28∘S

29∘S

30∘S

31∘S

32∘S

33∘S

34∘S

27∘S

28∘S

29∘S

30∘S

31∘S

32∘S

33∘S

34∘S

58∘ W

57∘ W

56∘ W

55∘ W

54∘ W

53∘ W

52∘ W

51∘ W

50∘ W

49∘ W

58∘ W

57∘ W

56∘ W

55∘ W

54∘ W

53∘ W

52∘ W

51∘ W

50∘ W

49∘ W

Brazil

Rio Grande do Sul state

Mariana PimentelStudy area A

Caccedilapava do SulStudy area B

Figure 3 Localization of the study areas in Mariana Pimentel (A) and Cacapava do Sul (B)

Figure 4 Study area A Morro do Papaleo outcrop Rio BonitoFormation From the base to topwe can identify the diamictite (lightgray bottom part) the carbonaceous pelite (black in the middlepart) and the sandstone (light and medium brown top part)

51 Ilris 3D Laser Scanner The Ilris 3D terrestrial laserscanner developed by the Canadian company Optech hasthe following technical characteristics described in Table 1(Optech 2009)

52 Definition of theWorkMethod The initial tests describedhere aimed to determine the ability of the TLS equipmentavailable to provide viable data for spatial and spectral pro-cessing and after reviewing the available data the method-ology of work to be used Some rock samples were placedon a bench located approximately 6m from the laser scannerdevice These samples were surveyed with a high resolutionof approximately 1mm between the points Among the rocksamples used were granite sandstone basalt carbonaceouspelite diamictite banded iron formation and gabbro

Figure 5 Study area B Pedra Pintada outcrop Pedra PintadaAlloformation The name of this outcrop has its origin due to thedifferent tones of reddish stratification levels The red color camefrom different concentrations of iron oxide present in the cement ofthe sandstones

Table 1 Ilris 3D terrestrial laser scanner technical characteristics

Range Up to 1200mMinimum range 3 metersLinear accuracy 7mm at 100mAngular accuracy 8mm at 100mBeam diameter 22 cm at 100mLaser repetition rate 2000 points per secondLaser class Class 1 (eye safe)Wavelength 1535 nm (infrared)Field of view 40∘ times 40∘

Digital camera Integrated 31M pixel

The first feature identified was the high noise whoserandom pattern did not allow the correct identification ofthe samples Subsequent tests proved that surveys carriedout with the Ilris 3D TLS in resolutions higher than 1 cmpresented noise due to the diameter of the emitted laser pulse

The second feature highlighted in the samples survey wasrelated to the intensity values of return from the rocks It wasexpected that the samples would exhibit completely differentspectral signatures which did not happenThis situation wasdue to the fact that the Ilris 3D equipment has been designedfor long distance surveys up to 1200meters on an object with


Figure 6 Initial survey with rock samples 60 meters apart fromthe scannerThe unusual magenta points around the samples are theedge effect

80 reflectance which in surveys in short distances makesthe laser pulse too intense giving a response that cannot beused to differentiate rocks

Works using intensity values of the laser pulse return asthe element of study (ie [42]) in general tend to maintaina minimum distance of approximately 60meters between theequipment and the surveyed target to avoid overexposure ofthe target This can represent a disadvantage in the case ofoutcrops situated in narrow canyons caves or other placeswhere the minimum distance from the equipment to thetarget is not possible

After the failure of processing the data gathered in theroom the equipment and samples were carried to the innercourtyard of the university keeping a distance of 60 metersbetween the equipment and the samples The samples wereplaced on a plastic holder and the survey was conductedIn this second test since the beginning of the survey it waspossible to see that at this distance the initial results obtainedattested the possibility of processing the intensity values of thesamples (Figure 6)

Another key component identified in the survey of thesamples was the presence of noise that presents a low graylevel near to the minimum values found in the point cloudaround the bracket assembly and samples This feature wasdue to the divergence of the laser which in a distance of 100meters covers a circle with 22 cm in diameter If we considerthat the spacing between points used for this surveywas 1mmand the average distance between the laser and the targets was60 meters we can infer that in the edges of the targets due tothe size of the laser pulse an increasingly smaller percentageof the pulse returns until the moment it ceases altogethercausing an area of low return (Figure 7)

53 Data Preparation The study areas A and B had alreadybeen surveyed previously in projects conducted by theGraduate Program in Geology in Unisinos and the datawere available for use in the LASERCA which simplified theprocesses of study and selection of the areas

1 cm

(a)

1 cm

(b)

1 cm

(c)

Figure 7 Edge effect in the targets using the Ilris 3D divergence(22 cm at 100m) and 1 cm space between the points (a) The laserpulse in full incidence on the target (b) The laser pulse with almost50 incidence on the target obtaining a second response from partof the pulse outside the target (c) The laser pulse with only 10 ofincidence in the target obtaining an extremely low intensity primarypoint and a stronger second response

Theuse of these data confirms a key feature for the surveysof point cloud acquisition the possibility of a historical recordof the outcrop and processing and interpretation even after itsweathering or destruction

The point clouds from both surveys were preprocessed inthe PARSER software in order to convert the Ilris 3D rawThe intensity data of the Morro do Papaleo outcrop wereexported as 16 gray levels and the Pedra Pintada outcrop with256 gray levels of intensity Despite the different ranges ofoutput intensity data in the outcrops they showed no useimpossibilities for the processing of return values

After the preprocessing of the point clouds they wereimported in the software PolyWorks (Inovx) specific soft-ware for processing data from TLS In PolyWorks all extra-neous elements were removed from the outcrop such asvegetation elements from anthropic action (garbage wastematerials etc) and any other elements that are not part ofthe features of the outcrop

The survey of the study area A Morro do Papaleo inthe city of Mariana Pimentel was performed with an averagespacing of 3 cm between the points obtained at an averagedistance of 48 meters Due to the conditions found at the


Figure 8 K-Clouds interfaceThe software allows the user to definethe number of clusters linked with the intensity values (I) or spatialvalues (XYZ) coordinates

site it was not possible to install the equipment over alonger distance However the data collected presented theright characteristics to be processed for use of its intensityA total of 8 stations were performed in the study area inorder to obtain the entirely outcrop but were selected an areaof approximately 36 square meters (6m times 6m) in an ideallocation where they were all major facies of the outcrop onsite

The survey of the study area B Pedra Pintada in the cityof Cacapava do Sul was performed with an average spacingof 4 cm at an average distance of approximately 232 metersbetween the equipment and the face of the outcrop Theconditions found at the site allowed the installation of theequipment at a distance suitable for the collection of datawhich allow processing of intensity values

At this location the main face of the outcrop has a lengthof approximately 274meters where we selected a smaller areawith a length of approximately 165 meters in order to allowgreater flexibility in the experiments and tests

The point clouds in both study areas were processed togenerate histograms of the spectral distribution of rocks inthe outcrops to assist in the process of choosing the numberof classes to be classified

54 K-Clouds Classification Program Currently the processof the intensity values of the laser scanner data is performedin conventional software for digital image processing Aimingto spread a specific method for use of TLS in outcrops aprogram was developed specifically to process point cloudsby applying the clustering classification algorithm 119896-means[40 43]

This software namedK-Clouds (Figure 8) was developedboth by V3D and UNISINOS It processes the TXT or PTSfile formats and perform the clustering of the intensity valuesconsidering the number of classes defined by the user

The information is exported through a file in XYZRGBformat where each class ldquo119896rdquo has a unique RGB encoding inorder to ease its identification This information can then beimported into software capable to handle point clouds Colorsare useful to enhance different spectral signatures relatedwithdifferent physical and chemical characteristics They allowthe identification of the rocks present in the outcrop and putin evidence characteristics that may complement the studiescarried out on site

Figure 9 Classified point cloud from the rock samples The basaltsample (cyanmdashclass 1) is easily recognizable as the edge effect aboveall samples but the other samples need a specific analysis to beidentified The granite is formed by 50 of class 2 (green) 40 ofclass 3 (purple) and 10 of class 4 (red) the banded iron formationis formed by 70 of class 2 15 of class 3 and 15 class 4 the gabbrois formed by 85 of class 4 and 15 class 2

6 Results and Discussion

In the initial testing phase with the samples it became clearthat the equipment used could not be operated so close tothe targets with the purpose of using the intensity values ofreturn due to its extremely strong laser pulse masking thespectral signature of the targetsThis issue has been solved bycollecting data at distances close to or greater than 50 meters

Another feature that could be provedwith this survey wasthe relationship betweendata andnoise foundAswe improvethe resolution of the survey reducing the space between thepoints we add a significant amount of noise points dislocatedfrom the actual position of the target Because of this featureand the difficulty of carrying out a data filtering the selectionof the proper space between the points becomes with theuse of an adequate distance between the equipment and thetarget a key element for a positive or negative work outcome

Data from the initial test were processed using the K-Clouds software and four (119896 = 4) distinct classes of rockswere selected one class for each rock sample usedThe resultsshow a large difference between the basalt sample (located tothe right) and the other samples (Figure 9)

Compared with the other samples the basalt due to itslow reflectivity and high capacity of energy absorption dueto its dark tint presented a low return of the laser pulseemittedThese featuresmade their identification easier by theclassifier

The external areas of the samples were subjected to theedge effect from the diameter of the laser and high resolutionproviding low intensity return values These points werewrongly placedwithin the same class of the basalt sampleTheother samples showed minor differences regarding a specificspectral signature compared with the sample of basalt but ifwe make an analysis of the amount of points from each class


0 2 4 6 8 10 12 14 16Gray level-return pulse intensity

0100020003000400050006000700080009000

Popu

latio

n

K1

K2

K3K4

K5

K6

Histogram of Morro do Papaleo

Figure 10 Histogram from the return intensity values from theMorro doPapaleo outcropThe red vertical lines separate the clustersthat were user-defined and applied in the K-Clouds software

in the samples we can differentiate the rocks The granite isformed by 50 of class 2 (green) 40 of class 3 (purple) and10 of class 4 (red) the banded iron formation is formed by70 of class 2 15 of class 3 and 15 of class 4 the gabbro isformed by 85 of class 4 and 15 of class 2

In the study area A Morro do Papaleo outcrop theresulting point cloud with about 51000 points was processedto generate a histogram of the spectral distribution from therocks available in the outcrop (Figure 10)

After the histogram analysis the point clouds wereclassified into six classes of intensity return (119896 = 6) based onthe spectral distribution of the intensity data (Figure 11)

In a direct visual comparison of the classified point cloudwith the reference image of the site it can be concludedthat the classifier was able to identify and separate thecarbonaceous pelite layer in the central portion of the outcrop(cyan) Above the carbonaceous pelite layer in the upperportion of the outcrop the existing sandstone was separatedinto two classes the colors dark blue and green The partsplaced in the dark blue exhibit a tendency to be sandstonewith higher degree of weathering while the parts placedin the green tend to be sandstone with lower degree ofweathering

At the bottom of the carbonaceous pelite the programidentified three distinct features in the diamictite which hasa tendency to be a less weathered rock near the carbonaceouspelite (red) and a more weathered rock near the base layer(magenta and yellow) It is also evident that in the lowerright portion of the study area a small block of carbonaceouspelite detached from the main layer which was successfullyidentified (cyan)

In the study area B the Pedra Pintada outcrop withapproximately 26 million points was processed in the samemanner as the study area A to generate a histogram of thespectral distribution of the existing rocks in the outcrop(Figure 12)

After the histogram analysis it was defined that the pointcloud should be classified into three classes of intensity return(119896 = 3) based on the spectral distribution of the intensitydata (Figure 13) In both cases along with the data from thehistogram a photo from the site assisted in determining thenumber of classes to be used

The use of only three classes of information in thisoutcrop in addition to the analysis of the histogram is based

on the separation of the different areas from the outcropconcerning the iron oxide concentration present which isclearly characterized by the different levels of reddish huepresent in the outcrop

In the classified point cloud it is possible to identifythe depositional structures enhanced by greater or lesserconcentration of iron oxide as cementThe green points referto higher concentrations of iron oxide and the red area refersto lower concentrations of this compound in the cement ofthe stone

Another factor to consider is that due to thewavelength of1535 nanometers from the Ilris 3D in the mid-infrared rangethere is a small penetration of the laser in areas where the rockis affected by lichens or thin weathering crusts Thereforethe spectral classification is not affected by such elementsThe comparison of the photo with the spectral classificationclearly shows that the sedimentary structures of the outcropare visible even in areas with lower quality of exposure

These type of features are very common in places withhigh temperature and humidity like those regions where testswere performed To improve data acquisition procedures andprocessing techniques are fundamental to better represent theoutcrops as digital models

7 Conclusions

Theuse of terrestrial laser scanner for applications in geologyand Digital Outcrop Modeling is a technological trendthat increasingly sees its place among the various classicalgeologic techniques already established Modern equipmentwith longer range higher quality better spatial accuracy ofthe data and reduced time for collection of data make theuse of terrestrial laser scanner a viable and reliable tool in theexecution of works

The classification of spectral patterns of intensity fromthe laser pulse return for geological applications is an areathat needs more developments and specific research aimedto establish and consolidate work methods In this contextit is clear that the development of the K-Clouds softwarethrough the use of 119896-means cluster classifier provides asimple and practical tool to be inserted within the workflowgeology professionals However other areas that use datafrom terrestrial laser scanner can use these methods

The application of this program to classify the intensityvalues of point clouds from outcrops presented consistentresults being able to identify differences between severalrocks presented in outcrops or even changes in chemicalcomposition related to the presence of cementwith iron oxidefound in specific levels of the Pedra Pintada outcrop

This algorithm is suitable for this type of applicationbeing capable to distinguish different types of rocks (Figures9 and 11) or a same rock with different chemical characteristic(Figures 11 and 13)

One of the improvements planned for the program is theuse of other classifiers in addition to the 119896-means They willprovide new results for different rocks such as carbonatesand evaporites Another element to be developed in the userinterface is the automatic display of the histogram from


SandstoneSandstoneCarbonaceous pelite

DiamictiteDiamictiteDiamictite

(a) (b)

Figure 11 Comparison between the classified point cloud (a) and the reference image from the outcrop (b)

0 50 100 150 200 250Gray level-return pulse intensity

010000200003000040000500006000070000

Popu

latio

n

K1

K2

K3

Histogram of Pedra Pintada

Figure 12 Histogram from the return intensity values from thePedra Pintada outcrop The red vertical lines separate the clustersthat were user-defined and applied in the K-Clouds software

the intensity of returned laser pulse in order to ease theuser understanding and simplify the process of choosing thecorrect number of user-defined classes to improve the resultsobtained by the classifier

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

The authors would like to thank the Private GraduateProgram (PROSUP) of the Bureau for the Qualificationof Higher Education Students (CAPES) This project wasfinancially supported by FAPERGS project ldquoModelosDigitais de Afloramentos como ferramenta na analisee interpretacao geologicardquo (ARDmdashProcess 100477-0) PETROBRAS through the projects NEAP (16-SAP4600242459) and ldquoMapeamento 3D Georreferenciado

(a)

Sandstone with lower content of iron oxide in the cementSandstone with higher content of iron oxide in the cement

(b)

Figure 13 Comparison between the classified point cloud (b) andthe reference image from the outcrop (a)

de Afloramentos Utilizando uma Tecnica LIDARrdquo (00500044869084-SAP 4600285973) and FINEP (Financiadorade Estudos e Projetos) project ldquoMODAmdashModelagemDigitalde Afloramentos Usando GPUrdquo (MCTFINEPmdashPre-SalCooperativos ICTmdashEmpresas 032010)

References

[1] S J Buckley J A Howell H D Enge and T H Kurz ldquoTerres-trial laser scanning in geology data acquisition processing andaccuracy considerationsrdquo Journal of the Geological Society vol165 no 3 pp 625ndash638 2008

[2] M J Arnot T R Good and J J M Lewis ldquoPhotogeological andimage-analysis techniques for collection of large-scale outcropdatardquo Journal of Sedimentary Research vol 67 no 5 pp 984ndash987 1997

[3] L Maerten D D Pollard and F Maerten ldquoDigital mappingof three-dimensional structures of the Chimney Rock fault


systems central Utahrdquo Journal of Structural Geology vol 23 no4 pp 585ndash592 2001

[4] X Xu C L V Aiken J P Bhattacharya et al ldquoCreating virtual3-D outcroprdquo Leading Edge vol 19 no 2 pp 197ndash202 2000

[5] X Xu J B Bhattacharya R K Davies and C L V AitkenldquoDigital geologic mapping of the Ferron sandstone MuddyCreek Utah withGPS and reflectorless laser rangefindersrdquoGPSSolutions vol 5 pp 15ndash23 2001

[6] M Alfarhan LWhite D Tuck and C Aiken ldquoLaser rangefind-ers and ArcGIS combined with three-dimensional photore-alistic modeling for mapping outcrops in the Slick HillsOklahomardquo Geosphere vol 4 no 3 pp 576ndash587 2008

[7] J A Bellian D C Jennette C Kerans et al ldquo3-Dimensional digital outcrop data collection and analysisusing eye-safe laser (LIDAR) technologyrdquo 2002 httpwwwsearchanddiscoverynetdocumentsbeg3dindexhtm

[8] J A Bellian C Kerans and D C Jennette ldquoDigital outcropmodels applications of terrestrial scanning lidar technology instratigraphic modelingrdquo Journal of Sedimentary Research vol75 no 2 pp 166ndash176 2005

[9] H D Enge S J Buckley A Rotevatn and J A HowellldquoFrom outcrop to reservoir simulation model workflow andproceduresrdquo Geosphere vol 3 no 6 pp 469ndash490 2007

[10] R R Jones K J W McCaffrey J Imber et al ldquoCalibrationand validation of reservoir models the importance of highresolution quantitative outcrop analoguesrdquo Geological SocietySpecial Publication vol 309 no 1 pp 87ndash98 2008

[11] P D White and R R Jones ldquoA cost-efficient solution to truecolor terrestrial laser scanningrdquoGeosphere vol 4 no 3 pp 564ndash575 2008

[12] J K Pringle A RWesterman J D Clark N J Drinkwater andA R Gardiner ldquo3D high-resolution digital models of outcropanalogue study sites to constrain reservoir model uncertaintyan example from Alport Castles Derbyshire UKrdquo PetroleumGeoscience vol 10 no 4 pp 343ndash352 2004

[13] R M Phelps and C Kerans ldquoArchitectural characterizationand three-dimensional modeling of a carbonate channel-leveecomplex permian SanAndres Formation Last ChanceCanyonNew Mexico USArdquo Journal of Sedimentary Research vol 77no 11-12 pp 939ndash964 2007

[14] I Fabuel-Perez D Hodgetts and J Redfern ldquoA new approachfor outcrop characterization and geostatistical analysis of a low-sinuosity fluvial-dominated succession using digital outcropmodels upper triassic oukaimeden sandstone formation cen-tral high Atlas Moroccordquo American Association of PetroleumGeologists Bulletin vol 93 no 6 pp 795ndash827 2009

[15] D Kurtzman J A El Azzi F J Lucia J Bellian C Zahmand X Janson ldquoImproving fractured carbonate-reservoir char-acterization with remote sensing of beds fractures and vugsrdquoGeosphere vol 5 no 2 pp 126ndash139 2009

[16] A Rotevatn S J Buckley J A Howell and H FossenldquoOverlapping faults and their effect on fluid flow in differentreservoir types a LIDAR-based outcrop modeling and flowsimulation studyrdquo American Association of Petroleum GeologistsBulletin vol 93 no 3 pp 407ndash427 2009

[17] K Verwer M-T Oscar J A M Kenter and G D PortaldquoEvolution of a high-relief carbonate platform slope using 3Ddigital outcrop models lower jurassic djebel bou dahar highatlas Moroccordquo Journal of Sedimentary Research vol 79 no 5-6 pp 416ndash439 2009

[18] H D Enge and J A Howell ldquoImpact of deltaic clinothemson reservoir performance dynamic studies of reservoir analogs

from the Ferron SandstoneMember andPantherTongueUtahrdquoAmerican Association of Petroleum Geologists Bulletin vol 94no 2 pp 139ndash161 2010

[19] I Fabuel-Perez D Hodgetts and J Redfern ldquoIntegrationof digital outcrop models (DOMs) and high resolutionsedimentologymdashworkflow and implications for geologicalmodelling Oukaimeden Sandstone Formation High Atlas(Morocco)rdquo Petroleum Geoscience vol 16 no 2 pp 133ndash1542010

[20] J A Bellian R Beck and C Kerans ldquoAnalysis of hyperspectraland lidar data remote optical mineralogy and fracture identifi-cationrdquo Geosphere vol 3 no 6 pp 491ndash500 2007

[21] M I Olariu J F Ferguson C L V Aiken and X Xu ldquoOut-crop fracture characterization using terrestrial laser scannersdeep-water Jackfork sandstone at Big Rock Quarry ArkansasrdquoGeosphere vol 4 no 1 pp 247ndash259 2008

[22] R R Jones S Kokkalas and K J W McCaffrey ldquoQuantitativeanalysis and visualization of nonplanar fault surfaces usingterrestrial laser scanning (LIDAR)-the Arkitsa fault centralGreece as a case studyrdquo Geosphere vol 5 no 6 pp 465ndash4822009

[23] C K Zahm and P H Hennings ldquoComplex fracture devel-opment related to stratigraphic architecture challenges forstructural deformation prediction Tensleep Sandstone at theAlcova anticlineWyomingrdquoAmerican Association of PetroleumGeologists Bulletin vol 93 no 11 pp 1427ndash1446 2009

[24] ANagalli A P Fiori andBNagalli ldquoMetodo paraAplicacao deEscaner a Laser Terrestre ao Estudo da Estabilidade de Taludesem Rochardquo Revista Brasileira de Geociencias vol 41 no 1 pp56ndash67 2011

[25] T F Wawrzyniec L D McFadden A Ellwein et al ldquoChrono-topographic analysis directly from point-cloud data a methodfor detecting small seasonal hillslope change Black MesaEscarpment NE Arizonardquo Geosphere vol 3 no 6 pp 550ndash5672007

[26] X Janson C Kerans J A Bellian and W Fitchen ldquoThree-dimensional geological and synthetic seismic model of EarlyPermian redeposited basinal carbonate deposits VictorioCanyon west Texasrdquo American Association of Petroleum Geol-ogists Bulletin vol 91 no 10 pp 1405ndash1436 2007

[27] K T Bates F Parity P L Manning et al ldquoHigh-resolutionLiDAR and photogrammetric survey of the Fumanya dinosaurtracksites (Catalonia) implications for the conservation andinterpretation of geological heritage sitesrdquo Journal of the Geo-logical Society vol 165 no 1 pp 115ndash127 2008

[28] C E Nelson D A Jerram R W Hobbs R Terrington andH Kessler ldquoReconstructing flood basalt lava flows in threedimensions using terrestrial laser scanningrdquo Geosphere vol 7no 1 pp 87ndash96 2011

[29] K J W McCaffrey M Feely R Hennessy and J ThompsonldquoVisualization of folding in marble outcrops Connemarawestern Ireland an application of virtual outcrop technologyrdquoGeosphere vol 4 no 3 pp 588ndash599 2008

[30] K J W McCaffrey D Hodgetts J Howell et al ldquoVirtualfieldtrips for petroleum geoscientistsrdquo Petroleum Geology Con-ference Series vol 7 pp 19ndash26 2010

[31] F Ferrari M R Veronez F M W Tognoli L C Inocencio PS G Paim and R M da Silva ldquoVisualizacao e interpretacaode modelos digitais de afloramentos utilizando laser scannerterrestrerdquo Geociencias UNESP vol 31 no 1 pp 79ndash91 2012

[32] G Petrie and C K Toth ldquoTerrestrial laser scannersrdquo in Topo-graphic Laser Ranging and Scanning Principles and Processing


J Shan and C K Toth Eds CRC Press Boca Raton Fla USA2009

[33] P Hariharan Basics of Interferometry Elsevier 2nd edition2007

[34] J I San Jose Alonso J M Rubio J J M Martin and J GFernandez ldquoComparing time-of-flight and phase-shiftrdquo in Thesurvey of the Royal Pantheon in the Basilica of San Isidoro (Leon)ISPRS Workshop ldquo3D-ARCH 2011rdquo 3D Virtual Reconstructionand Visualization of Complex Architectures Trento ItalyMarch 2011

[35] B Ergun ldquoTerrestrial laser scanner data integration in survey-ing engineeringrdquo in Laser ScanningTheory and Applications CWang Ed InTech 2011

[36] B Mirkin Mathematical Classification and Clustering KluwerAcademic Publishers 1996

[37] N Brodu and D Lague ldquo3D terrestrial lidar data classificationof complex natural scenes using a multi-scale dimensionalitycriterion applications in geomorphologyrdquo ISPRS Journal ofPhotogrammetry and Remote Sensing vol 68 no 1 pp 121ndash1342012

[38] M Franceschi G Teza N Preto A Pesci A Galgaro and SGirardi ldquoDiscrimination between marls and limestones usingintensity data from terrestrial laser scannerrdquo ISPRS Journal ofPhotogrammetry andRemote Sensing vol 64 no 6 pp 522ndash5282009

[39] G Gan C Ma and J Wu Data Clustering Theory Algorithmsand Applications SIAM Philadelphia Pa USA ASA Alexan-dria Va USA 2007

[40] J B MacQueen ldquoSome methods for classification and analysisof multivariate observationsrdquo in Proceedings of 5th BerkeleySymposium on Mathematical Statistics and Probability vol 1pp 281ndash297 University of California Press Berkeley Calif USA1967

[41] P S G Paim C Fallgatter and A S Silveira ldquoGuaritasdo Camaqua RSmdashexuberante cenario com formacoesgeologicas de grande interesse didatico e turısticordquo inSıtios Geologicos e Paleontologicos do Brasil M WingeC Schobbenhaus C R G Souza et al Eds 2010httpwwwunbbrigsigepsitio076sitio076pdf

[42] D Burton D B Dunlap L J Wood and P P Flaig ldquoLidarintensity as a remote sensor of rock propertiesrdquo Journal ofSedimentary Research vol 81 no 5 pp 339ndash347 2011

[43] K Alsabti S Ranka and V Singh ldquoAn efficient K-meansclustering algorithmrdquo in Proceedings of the 1stWorkshop onHighPerformance Data Mining Orlando Fla USA 1998

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ClimatologyJournal of

EcologyInternational Journal of


EarthquakesJournal of


Hindawi Publishing Corporationhttpwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2014

Mining


Journal of

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

International Journal of

Geophysics

OceanographyInternational Journal of


Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofPetroleum Engineering


GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Atmospheric SciencesInternational Journal of


OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Advances in


MineralogyInternational Journal of


MeteorologyAdvances in

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Geological ResearchJournal of


Geology Advances in


that is the spacing between the points collected at a certaindistance and high degree of accuracy of the survey

The use of TLS in outcrop studies is expanding due tothe ease of acquisition of georeferenced data in an accuratefast and automated way Its use for this purpose beganabout a decade ago [7] but only in recent years the numberof scientific papers has increased significantly Topics ofinterest are varied with emphasis on the methodologicalapproaches [1 8ndash11] reservoirs analogues studies [12ndash19] offractured rocks [20ndash23] slope stability [24] erosion rates[25] synthetic seismic models [26] geological heritage [27]determining the direction of basaltic lava flow [28] geologyeducation [29 30] and outcrop interpretation [31]

The equipment available on the market is able to conductthe survey and collection of thousands (time of flight) upto millions (phase) of points per second and some have theability to record the intensity of the pulse return for eachposition of the laser XYZ [8] The intensity of the returnpulse from the laser and the equipment range are directlyrelated to factors that affect theworking of the systemAmongthese factors we can mention the power of the emitted pulseand the reflectivity of the targets

The relation between the reflectivity of the targets and theintensity of the laser beam emitted is directly proportional tothe range of the Laser Scanner If a target has 10 reflectivitythe equipment should be closer than a target that has 90reflectivity [32]

As other active remote sensing systems the TLS has onlyone wavelength as the intensity of the return pulse is basedon the interaction of the laserrsquos specific wavelength with thetargetThis interactionmay be analyzed using interferometrytechniques which are the study of phenomena caused by theinterfering within the light waves which in this case are theinteraction between the laser wavelength and the physical-chemical characteristics of the targets [33]

As we use TLS to survey an outcrop we not only collecttheir spatial characteristics such as features and volumesbut we also obtain additional data that may allow a greaterunderstanding of the outcrop

In most cases each TLS equipment generates its own fileformat for archiving of the collected data However these fileshave similarities concerning the information stored Thesefiles record the spatial position of the points (X Y and Z) theintensity of the laser pulse that returns from the target (i) andthe RGB color pattern obtained from the pictures recordedby the internal camera The output is a txt file that providesthe sequence X Y Z I R G B

The spatial information is commonly used for volumecalculations and identification of geometric features and theRGB data assist in identifying visual features and provide abetter understanding of the objects surveyed with the TLSbut when it comes to the intensity values of the returnlaser pulse there is a lack of specific applications designedto transform this data in information which can detectvariations in the physical-chemical characteristics of thetargets

Thus the main objective of this study was to developa method for classification of spectral patterns in LIDAR

(light detect and range) data for the identification of rocks inoutcrops with distinct lithology

The validation of this method occurred through the useof a cluster analysis algorithm known as k-meansThe choiceof this algorithm was given by the obtained separation ofthe spectral responses of rocks with different characteristicsboth chemical andor physical This allowed us to classifythe patterns of spectral response and thus identify rocks andfeatures present in the outcrops This classifier tool will beincorporated in the software Mountain View (partnershipV3DUNISINOS) which allows the user view edits andprocesses data from TLS and the main application is focusedon Digital Outcrop Modeling (DOM)

2 Terrestrial Laser Scanner

The Laser Scanning and profiling systems in this caseterrestrial basically consist of a measuring instrument basedon laser that can measure vertical angles horizontal anglesand distances with a high standard of accuracy and speed bymeans of a mobile mirrors or prisms system that allow themapping of topographical features on targets

These systems have different methods for measure dis-tances and the time of flight (TOF) (Figure 1) and phase shift(PS) are the most widespread In the Time of Flight methodthe instrument sends a high-intensity laser pulse toward thetarget and measure the time between sending and receivingthe pulse returns and divides the value by 2 to calculate thedistance

In the phase shift method the instrument emanates alaser pulse fitted within a harmonic wave and the distance iscalculated by comparing the difference of the transmitted andreceived waves [35] (Figure 2)

Both methods of data acquisition have strengths andweaknesses due to the method applied The equipment thatuses the phase shift technology tends to collect more pointsper second at the same time interval millions of points persecond while the time of flight collects thousands of pointsper second However the phase shift scanners have a reducedrange that varies between 80 to 120 meters against up totwo kilometers in some time of flight equipment Anotherelement to consider is that the operation of phase shiftequipment in conditions of high insolation can reduce therange of the equipment

All the elements quoted above must be taken into con-sideration when selecting the most suitable technology toperform the work in which case Digital Outcrop Modeling

3 Classification

The term classification is defined as an actual or idealarrangement relating what is similar and separating what isdifferent having as main objectives to maintain and buildknowledge and analyze and relate the structure of certainphenomena [36] If we apply this definition to geologyspecifically in outcrops classification it becomes clear thatthe first and main elements to be classified and identified arethe rocks on site This information is intended to expand the


Emittedpulse

Primaryreturn pulse


Time

Inte

nsity








x(t)

A

T

t



4 Study Areas







27∘S

28∘S

29∘S

30∘S

31∘S

32∘S

33∘S

34∘S

27∘S

28∘S

29∘S

30∘S

31∘S

32∘S

33∘S

34∘S

58∘ W

57∘ W

56∘ W

55∘ W

54∘ W

53∘ W

52∘ W

51∘ W

50∘ W

49∘ W

58∘ W

57∘ W

56∘ W

55∘ W

54∘ W

53∘ W

52∘ W

51∘ W

50∘ W

49∘ W

Brazil





















1 cm

(a)

1 cm

(b)

1 cm

(c)
























0100020003000400050006000700080009000

Popu

latio

n

K1

K2

K3K4

K5

K6















7 Conclusions









(a) (b)



010000200003000040000500006000070000

Popu

latio

n

K1

K2

K3






Acknowledgments


(a)


(b)



References


























































Volume 2014

Mining


Journal of



Geophysics







Journal of




Advances in











Geology Advances in


Emittedpulse

Primaryreturn pulse


Time

Inte

nsity








x(t)

A

T

t



4 Study Areas







27∘S

28∘S

29∘S

30∘S

31∘S

32∘S

33∘S

34∘S

27∘S

28∘S

29∘S

30∘S

31∘S

32∘S

33∘S

34∘S

58∘ W

57∘ W

56∘ W

55∘ W

54∘ W

53∘ W

52∘ W

51∘ W

50∘ W

49∘ W

58∘ W

57∘ W

56∘ W

55∘ W

54∘ W

53∘ W

52∘ W

51∘ W

50∘ W

49∘ W

Brazil





















1 cm

(a)

1 cm

(b)

1 cm

(c)
























0100020003000400050006000700080009000

Popu

latio

n

K1

K2

K3K4

K5

K6















7 Conclusions









(a) (b)



010000200003000040000500006000070000

Popu

latio

n

K1

K2

K3






Acknowledgments


(a)


(b)



References


























































Volume 2014

Mining


Journal of



Geophysics







Journal of




Advances in











Geology Advances in


27∘S

28∘S

29∘S

30∘S

31∘S

32∘S

33∘S

34∘S

27∘S

28∘S

29∘S

30∘S

31∘S

32∘S

33∘S

34∘S

58∘ W

57∘ W

56∘ W

55∘ W

54∘ W

53∘ W

52∘ W

51∘ W

50∘ W

49∘ W

58∘ W

57∘ W

56∘ W

55∘ W

54∘ W

53∘ W

52∘ W

51∘ W

50∘ W

49∘ W

Brazil





















1 cm

(a)

1 cm

(b)

1 cm

(c)
























0100020003000400050006000700080009000

Popu

latio

n

K1

K2

K3K4

K5

K6















7 Conclusions









(a) (b)



010000200003000040000500006000070000

Popu

latio

n

K1

K2

K3






Acknowledgments


(a)


(b)



References


























































Volume 2014

Mining


Journal of



Geophysics







Journal of




Advances in











Geology Advances in








1 cm

(a)

1 cm

(b)

1 cm

(c)
























0100020003000400050006000700080009000

Popu

latio

n

K1

K2

K3K4

K5

K6















7 Conclusions









(a) (b)



010000200003000040000500006000070000

Popu

latio

n

K1

K2

K3






Acknowledgments


(a)


(b)



References


























































Volume 2014

Mining


Journal of



Geophysics







Journal of




Advances in











Geology Advances in



















0100020003000400050006000700080009000

Popu

latio

n

K1

K2

K3K4

K5

K6















7 Conclusions









(a) (b)



010000200003000040000500006000070000

Popu

latio

n

K1

K2

K3






Acknowledgments


(a)


(b)



References


























































Volume 2014

Mining


Journal of



Geophysics







Journal of




Advances in











Geology Advances in



0100020003000400050006000700080009000

Popu

latio

n

K1

K2

K3K4

K5

K6















7 Conclusions









(a) (b)



010000200003000040000500006000070000

Popu

latio

n

K1

K2

K3






Acknowledgments


(a)


(b)



References


























































Volume 2014

Mining


Journal of



Geophysics







Journal of




Advances in











Geology Advances in




(a) (b)



010000200003000040000500006000070000

Popu

latio

n

K1

K2

K3






Acknowledgments


(a)


(b)



References


























































Volume 2014

Mining


Journal of



Geophysics







Journal of




Advances in











Geology Advances in























































Volume 2014

Mining


Journal of



Geophysics







Journal of




Advances in











Geology Advances in























Volume 2014

Mining


Journal of



Geophysics







Journal of




Advances in











Geology Advances in










Volume 2014

Mining


Journal of



Geophysics







Journal of




Advances in











Geology Advances in

Documents

Research Article Spectral Pattern Classification in …downloads.hindawi.com › journals › tswj › 2014 › 539029.pdfCac¸apava do Sul, approximately km from Porto Alegre with