Page No. VI List of Figures VII List of Tables VIII List of Abbreviations Ix List of Symbols X 1. 1-12
Page No. VI List of Figures VII List of Tables VIII List of Abbreviations Ix List of Symbols X 1. 1-12
Page No.
Abstract vi
List of Figures vii
List of Tables viii
List of Abbreviations ix
List of Symbols x
1. Introduction 1-12
1.1. Motivation and Background 1
1.2. Data Mining 1-2
1.3. Introduction to Remote Sensing 2-3
1.4. Remote Sensing Process 3-4
1.5. Satellite Image 4
1.6. Satellite Communication 4-5
1.7. Digital Number 5-6
1.8. Image Classification 6
1.8.1. Unsupervised Classification 6-7
1.8.2. Supervised Classification 7-8
1.9. Image Classification Techniques 8
1.10. Traditional Image Classification Techniques 8
1.11. Parellelopiped Classification 8-9
1.12. Minimum Distance to Mean Classification 9
1.13. Maximum Likelihood Classification 10
1.14. Recent Soft Computing Techniques 10
1.15. Rough Set Classification 10-11
1.16. Fuzzy Set Classification 11-12
2. Literature Review 13-17
2.1. Literature Survey 13-16
2.2. Formulation of Problem 16
2.3. Objective 17
3. Methodology 18-26
3.1. Basic Design of Proposed Work 18
3.2. Input Images to be Classified 18-22
3.3. Description of Proposed Algorithm 23
3.4. Training Set Generation 24
iv
3.5. Rule Set Generation 24
3.6. Region Matching/ Classification 24
3.7. Accuracy Assessment 24
3.8. Classification Error Matrix 25-26
4. Results & Discussions 27-34
4.1. Experimental Study 27
4.2. Training Set Generation 28
4.3. Identification Process 28
4.4. Input Image 28
4.5. Simulation Interface 29-33
4.6. Extracting Terrain Feature from the given Image 34
4.7. Comparative Analysis 34
5. Conclusion & Future Scope 35-36
5.1. Conclusion 35-36
5.2. Future Scope 36
Publications 37
References 38-40
Appendices 41-58
Appendix A (Nomenclature) 41
Appendix B (Glossary) 42
Appendix C (An introduction to MATLAB) 43-44
Appendix D (MATLAB Code) 45-57
Appendix E (An introduction to ERDAS) 58
v
ABSTRACT
Land cover mapping is a pattern classification problem. Extensive research has been carried
out to use various computational techniques for remote sensing applications. Computational
intelligence components have been applied individually to solve the problem of Satellite
Image classification. Neural networks have been used extensively for the image classification
and have been reported to perform better when compared with the Statistical Classifiers.
Fuzzy set and Rough set theory have also been beneficial for the image classification.
In addition to these algorithms, there is a new range of computational algorithms that have
emerged from the behaviour of social insects called as Swarm Intelligence. Generally
speaking, swarm intelligence denotes that more complex intelligence emerges from the
interactions of individuals who behave as a swarm. The two most widely used swarm
intelligence approaches are Ant Colony Optimization (ACO) and Particle Swarm
Optimization (PSO).
The aim of this work is to propose a branch of natural computation i.e. Membrane computing
for image classification. The membrane computing is integrated itself for classifying an
image. With the extensive use of images classification for the Land Cover Mapping it is
required that the approach used for the image classification should be more accurate and
efficient. Image recognition techniques facilitate the classification of images into
semantically-meaningful categories and then map the land by the labels that have been
manually assigned to the categories. Automatic image classification is the task of classifying
images into semantic categories with or without supervised training.
As the traditional techniques for the automatic Image Classification have their certain
shortcomings like high resolution images needed for better information retrieval and the
accuracy limitations are also there. Also the recent Soft Computing approaches for Image
Classification are not able to provide good results in case of ambiguity. So for achieving the
better accuracy even with the low resolution satellite images and better land cover mapping,
the objective is to use Swarm Intelligence for Membrane Computing Image Classification.
Here we are using the algorithm of Swarm intelligence i.e. MC for the Satellite Image
Classification, so that we can retrieve the more accurate information about any Land Area
even with the low resolution satellite images. This approach is used because it provides the
greater speed and accuracy in its computation. As outcome it comes with 5% of kappa
improvement.
vi
LIST OF FIGURES
Page No.
Figure 1.1 Basic Remote Sensing Process 4
Figure 1.2 Supervised Classification 8
Figure 1.3 Minimum Distance to Mean Classification 9
Figure 1.4 Concept of Rough Set 11
Figure 3.1 Flow Chart of Proposed Work 18
Figure 3.2 False Color Composition 19
Figure 3.3 Image-1 19
Figure 3.4 Image-2 20
Figure 3.5 Image-3 20
Figure 3.6 Image-4 21
Figure 3.7 Image-5 21
Figure 3.8 Image-6 22
Figure 3.9 Image-7 22
Figure 3.10 Diagrammatic Representation of the Proposed Algorithm 23
Figure 4.1 Band images of Alwar area 27
Figure 4.2 Training Data Set for Alwar Area 28
Figure 4.3 Input image 28
Figure 4.4 GUI for Classification 29
Figure 4.5 Original Raw image 29
Figure 4.6 Classification 30
Figure 4.7 Comparative Analysis 30
Figure 4.8 Classified Rocky Image 31
Figure 4.9 Classified Urban Image 31
Figure 4.10 Classified Vegetation Image 32
Figure 4.11 Classified Water Image 32
Figure 4.12 Classified Barren Image 33
Figure 4.13 Complete Simulation 33
Figure 4.14 Extracted Image 34
vii
LIST OF TABLES
Page No.
Table 3.1 Details of the images to be classified 18
Table 3.2 Error Matrix 26
Table 4.1 Comparative analysis 34
viii
LIST OF ABBREVIATIONS
ix
LIST OF SYMBOLS
λ Wavelength
υ Frequency
r No. of rows in the error matrix
xii No. of observations in row i and column i (on the major diagonal)
xi+ Total of observations in row i (shown as marginal total to right of the matrix)
x+i Total of observations in column i (shown as marginal total at bottom of the matrix)
N Total number of observations included in matrix
k KHAT statistic