14
GIS & REMOTE SENSING LABORATORY SESSION 6: SUPERVISED AND UNSUPERVISED CLASSIFICATION NAME : XXXXXXXXXXXXXXXXXXXXXXXXXXXX MATRIC NO. : UK XXXXX PROGRAM : XXXXXXXX DATE : 19 AUGUST XXXX

Gis And Remote Sensing Lab 6

  • Upload
    iser

  • View
    130

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Gis And Remote Sensing Lab 6

GIS & REMOTE SENSING

LABORATORY SESSION 6:

SUPERVISED AND UNSUPERVISED

CLASSIFICATION

NAME : XXXXXXXXXXXXXXXXXXXXXXXXXXXXMATRIC NO. : UK XXXXXPROGRAM : XXXXXXXXDATE : 19 AUGUST XXXX

Page 2: Gis And Remote Sensing Lab 6

1. INTRODUCTION

One of the main purposes of satellite remote sensing is to interpret the observed data and

classify features. In addition to the approach of photo interpretation, quantitative analysis,

which uses computer to label each pixel to particular spectral classes (called

classification), is commonly used. Quantitative analysis can perform true multispectral

analysis, make use of all the available brightness levels and obtain high quantitative

accuracy. There are two broads of classification procedures: supervised classification

unsupervised classification.

The supervised classification is the essential tool used for extracting quantitative

information from remotely sensed image data. Using this method, the analyst has

available sufficient known pixels to generate representative parameters for each class of

interest. This step is called training. Once trained, the classifier is then used to attach

labels to all the image pixels according to the trained parameters. The most commonly

used supervised classification is maximum likelihood classification (MLC), which

assumes that each spectral class can be described by a multivariate normal distribution.

Another broad of classification is unsupervised classification. It doesn’t require human to

have the foreknowledge of the classes, and mainly using some clustering algorithm to

classify an image data. These procedures can be used to determine the number and

location of the unimodal spectral classes.

2. OBJECTIVES

The objectives of this experiment are:

i. To expose student on the usage of PCI Geomatica

ii. To introduce the technique of classification using PCI software.

iii. To teach the student basic guideline to do training area.

iv. To introduce various methods of classification such as minimum distance

classifier, parallelepiped classifier and maximum likelihood.

Page 3: Gis And Remote Sensing Lab 6

3. METHODOLOGY

3.1 Unsupervised classification

1) Open Focus window> Load the image from folder > Click Clipping / Sub setting>

Browse> Save file name

2) Right click on the color composite map layer > Image Classification>

Unsupervised.

3) In Session Selection dialog > Click New Session > Session Configuration> Add

Layer button > Enter 5 in Channels to add > Click Add> Close

4) Session Configuration:

i. Specify the RGB false color > Channel 4 for red, Channel 5 for green,

Channel 3 for blue

ii. Select Input Channels > Channel 1 until 7 except 6

iii. Select Output Channel > Channel 9 > OK

5) Unsupervised Classification > Max Class value - 10 > Show report, Save

signatures and Create PCT > OK

Page 4: Gis And Remote Sensing Lab 6

6) Save the classification report.

7) The steps from 1 to 6 are repeated for 20 and 30 MAX Class respectively.

3.2 Supervised classification

1) Open Focus > Load image > Clipping / Subsetting> Browse > Save the file name

2) Right click on the color composite map layer > Image Classification >

Supervised

3) In Session Selection dialog > Click New Session > Session Configuration > Open

Add Layer button > Channels to add (enter 2) > Add > Close

4) Session Configuration:

i. Specify the RGB false color > Channel 4 for red, Channel 5 for green,

Channel 3 for blue

ii. Select Input Channels > Channel 1 until 7 except 6

iii. Select Training Channel> Channel 11

iv. Select Output Channel> Channel 12

Page 5: Gis And Remote Sensing Lab 6

5) In Training Site Editor> Open Class> Click New to add new class

6) In the Class Editing section of the Training Set Editing panel, Select shape >

Draw an outline / shape around the area to be classified.

- Double click to finish the line

- Click the colored box (in training site editor ) to change the color

7) Start the classification process> Right click on Classification MetaLayer shortcut

menu> Run classification

Page 6: Gis And Remote Sensing Lab 6

8) In classify window, select algorithm, maximum likelihood parameters, and

classification windows exactly as the image below > Click OK.

9) After the steps are finish > Click save then close.

10) Save the Classification report.

Page 7: Gis And Remote Sensing Lab 6

4. RESULTS

4.1 Unsupervised Classification

i. Unsupervised classification with 10 classs

ii. Unsupervised classification with 20 class

Page 8: Gis And Remote Sensing Lab 6

iii. Unsupervised classification with 30 classes

4.2 Supervised Classification

i. Result for supervised classification

Page 9: Gis And Remote Sensing Lab 6

ii. Classification report

Page 10: Gis And Remote Sensing Lab 6

5. DISCUSSION

Unsupervised classification is where groupings of pixels with common characteristics are

based on the software analysis of an image without the user providing sample classes.

During laboratory, user rely on computer to uses techniques to determine which pixels

are related and what classes belong together. However, the user must have knowledge of

the area being classified when the groupings of pixels with common characteristics

produced by the computer have to be related to actual features on the ground such as

wetlands, developed areas, forests, etc.

In this session, supervised classification was done based on the idea that a user can select

sample pixels in an image that are representative of specific classes and then direct PCI

Geomatica software to use these choices as references for the classification of all other

pixels in the image. Training areas are selected based on the knowledge of the user. Then,

user also sets the bounds for how close the matches have to be. These bounds are set

based on the spectral characteristics of the training area, based on brightness or strength

of reflection in specific spectral bands / color. The user also designates the outputs which

is 10 classes is selected for classification.

In this experiment, the size of the shape drawn to classify training area should be

suitable to avoid overlapping with other objects. Then, the pixel selected for each class

must be more than 200 to ensure good result. Band 6 is omitted from the session of

configuration, because it did not detect any reflection from the objects and it only detect

temperature change in environment. In this experiment, user achieved average accuracy

of 84.18% and overall accuracy of 97.21% on the classification of the chosen 10 classes

which are forest, ocean, swamp, river, mangrove, of young and mature oil palm, urban,

grass, and vegetation.

For the unsupervised classification, user had selected three different maximum

classes which are 10, 20 and 30 classes and let the software analyses the unknown pixels

in the photo and divides into a number of classed based on natural groupings present in

the image values. Then, the result shown that more group are classified (based on number

Page 11: Gis And Remote Sensing Lab 6

of color appeared) on the highest maximum class (30 classes) compared to the less

selected.

6. CONCLUSION

To be concluded, this experiment exposes students on the technique of classification of

objects which are supervised and unsupervised classification with the minimum aid from

the demonstrator. In this experiment, we used maximum likelihood classifier for

supervised classification with 10 different classes and unsupervised classification using

different maximum classes of 10, 20 and 30 respectively. Therefore, the objectives are

achieved.