Classification Methods for Remotely Sensed Data, Second Edition

by Mather; Paul

• ISBN13: 9781420090727
• ISBN10: 1420090720
• Edition: 2nd
• Format: Hardcover
• Copyright: 2009-05-12
• Publisher: CRC Press

Summary

The field of pattern recognition is expanding in new directions, for example, in data mining and Earth observation data processing. This book covers the entire field of classification methods applied to remotely sensed data. After an introduction to the basics of remote sensing, the text provides a detailed discussion of different approaches to image classification, including maximum likelihood, fuzzy sets, and artificial neural networks. The second edition features three new chapters that cover recent developments in support vector machines, decision trees, as well as change detection and mixed pixel classification. It also updates references and discussions of Earth observation missions.Features· Covers the entire field of classification methods applied to remotely sensed data· Presents recent developments in support vector machines and decision trees· Updates extensive bibliography· Includes new chapter on change detectionand mixed pixel classificationAudienceGraduate and doctoral students working with remote sensing data in digital image analysis and GIS environments.

Table of Contents

Preface to the Second Edition	p. xi
Preface to the First Edition	p. xiii
Author Biographies	p. xix
Remote Sensing in the Optical and Microwave Regions	p. 1
Introduction to Remote Sensing	p. 4
Atmospheric Interactions	p. 5
Surface Material Reflectance	p. 5
Spatial and Radiometric Resolution	p. 8
Optical Remote Sensing Systems	p. 10
Atmospheric Correction	p. 11
Dark Object Subtraction	p. 12
Modeling Techniques	p. 13
Modeling the Atmospheric Effect	p. 13
Steps in Atmospheric Correction	p. 17
Correction for Topographic Effects	p. 19
Remote Sensing in the Microwave Region	p. 22
Radar Fundamentals	p. 23
SLAR Image Resolution	p. 24
Geometric Effects on Radar Images	p. 26
Factors Affecting Radar Backscatter	p. 29
Surface Roughness	p. 29
Surface Conductivity	p. 30
Parameters of the Radar Equation	p. 30
Imaging Radar Polarimetry	p. 31
Radar Polarization State	p. 32
Polarization Synthesis	p. 34
Polarization Signatures	p. 35
Radar Speckle Suppression	p. 37
Multilook Processing	p. 37
Filters for Speckle Suppression	p. 38
Pattern Recognition Principles	p. 41
Feature Space Manipulation	p. 42
Tasseled Cap Transform	p. 45
Principal Components Analysis	p. 46
Minimum/Maximum Autocorrelation Factors (MAF)	p. 50
Maximum Noise Fraction Transformation	p. 51
Feature Selection	p. 52
Fundamental Pattern Recognition Techniques	p. 54
Unsupervised Methods	p. 54
The k-means Algorithm	p. 54
Fuzzy Clustering	p. 56
Supervised Methods	p. 57
Parallelepiped Methods	p. 57
Minimum Distance Classifier	p. 57
Maximum Likelihood Classifier	p. 58
Combining Classifiers	p. 61
Incorporation of Ancillary Information	p. 62
Use of Texture and Context	p. 63
Using Ancillary Multisource Data	p. 63
Sampling Scheme and Sample Size	p. 65
Sampling Scheme	p. 66
Sample Size, Scale, and Spatial Variability	p. 67
Adequacy of Training Data	p. 69
Estimation of Classification Accuracy	p. 69
Epilogue	p. 74
Artificial Neural Networks	p. 77
Multilayer Perceptron	p. 77
Back-Propagation	p. 78
Parameter Choice, Network Architecture, and Input/Output Coding	p. 82
Decision Boundaries in Feature Space	p. 84
Overtraining and Network Pruning	p. 88
Kohonen's Self-Organizing Feature Map	p. 90
SOM Network Construction and Training	p. 90
Unsupervised Training	p. 91
Supervised Training	p. 93
Examples of Self-Organization	p. 94
Counter-Propagation Networks	p. 98
Counter-Propagation Network Training	p. 99
Training Issues	p. 101
Hopfield Networks	p. 101
Hopfield Network Structure	p. 102
Hopfield Network Dynamics	p. 102
Network Convergence	p. 103
Issues Relating to Hopfield Networks	p. 105
Energy and Weight Coding: An Example	p. 106
Adaptive Resonance Theory (ART)	p. 108
Fundamentals of the ART Model	p. 109
Choice of Parameters	p. 112
Fuzzy ARTMAP	p. 113
Neural Networks in Remote Sensing Image Classification	p. 116
An Overview	p. 116
A Comparative Study	p. 119
Support Vector Machines	p. 125
Linear Classification	p. 126
The Separable Case	p. 126
The Nonseparable Case	p. 129
Nonlinear Classification and Kernel Functions	p. 130
Nonlinear SVMs	p. 130
Kernel Functions	p. 132
Parameter Determination	p. 135
t-Fold Cross-Validations	p. 137
Bound on Leave-One-Out Error	p. 138
Grid Search	p. 140
Gradient Descent Method	p. 142
Multiclass Classification	p. 144
One-against-One, One-against-Others, and DAG	p. 144
Multiclass SVMs	p. 146
Vapnik's Approach	p. 146
Methodology of Crammer and Singer	p. 147
Feature Selection	p. 149
SVM Classification of Remotely Sensed Data	p. 150
Concluding Remarks	p. 153
Methods Based on Fuzzy Set Theory	p. 155
Introduction to Fuzzy Set Theory	p. 155
Fuzzy Sets: Definition	p. 156
Fuzzy Set Operations	p. 157
Fuzzy C-Means Clustering Algorithm	p. 159
Fuzzy Maximum Likelihood Classification	p. 162
Fuzzy Rule Base	p. 164
Fuzzification	p. 165
Inference	p. 169
Defuzzification	p. 171
Image Classification Using Fuzzy Rules	p. 173
Introductory Methodology	p. 173
Experimental Results	p. 178
Decision Trees	p. 183
Feature Selection Measures for Tree Induction	p. 184
Information Gain	p. 185
Gini Impurity Index	p. 188
ID3, C4.5, and SEE5.0 Decision Trees	p. 189
ID3	p. 189
C4.5	p. 193
SEE5.0	p. 196
CHAID	p. 197
CART	p. 198
QUEST	p. 201
Split Point Selection	p. 201
Attribute Selection	p. 203
Tree Induction from Artificial Neural Networks	p. 204
Pruning Decision Trees	p. 205
Reduced Error Pruning (REP)	p. 207
Pessimistic Error Pruning (PEP)	p. 207
Error-Based Pruning (EBP)	p. 208
Cost Complexity Pruning (CCP)	p. 209
Minimal Error Pruning (MEP)	p. 212
Boosting and Random Forest	p. 214
Boosting	p. 214
Random Forest	p. 215
Decision Trees in Remotely Sensed Data Classification	p. 217
Concluding Remarks	p. 220
Texture Quantization	p. 221
Fractal Dimensions	p. 222
Introduction to Fractals	p. 223
Estimation of the Fractal Dimension	p. 224
Fractal Brownian Motion (FBM)	p. 225
Box-Counting Methods and Multifractal Dimension	p. 226
Frequency Domain Filtering	p. 231
Fourier Power Spectrum	p. 231
Wavelet Transform	p. 235
Gray-Level Co-Occurrence Matrix (GLCM)	p. 239
Introduction to the GLCM	p. 239
Texture Features Derived from the GLCM	p. 241
Multiplicative Autoregressive Random Fields	p. 243
MAR Model: Definition	p. 243
Estimation of the Parameters of the MAR Model	p. 245
The Semivariogram and Window Size Determination	p. 246
Experimental Analysis	p. 249
Test Image Generation	p. 249
Choice of Texture Features	p. 250
Multifractal Dimension	p. 250
Fourier Power Spectrum	p. 250
Wavelet Transform	p. 250
Gray-Level Co-Occurrence Matrix	p. 250
Multiplicative Autoregressive Random Field	p. 251
Segmentation Results	p. 251
Texture Measure of Remote Sensing Patterns	p. 252
Modeling Context Using Markov Random Fields	p. 255
Markov Random Fields and Gibbs Random Fields	p. 256
Markov Random Fields	p. 256
Gibbs Random Fields	p. 257
MRF-GRF Equivalence	p. 259
Simplified Form of MRF	p. 261
Generation of Texture Patterns Using MRF	p. 263
Posterior Energy for Image Classification	p. 264
Parameter Estimation	p. 267
Least Squares Fit Method	p. 268
Results of Parameter Estimations	p. 271
MAP-MRF Classification Algorithms	p. 273
Iterated Conditional Modes	p. 274
Simulated Annealing	p. 275
Maximizer of Posterior Marginals	p. 277
Experimental Results	p. 278
Multisource Classification	p. 283
Image Fusion	p. 284
Image Fusion Methods	p. 284
Assessment of Fused Image Quality in the Spectral Domain	p. 287
Performance Overview of Fusion Methods	p. 288
Multisource Classification Using the Stacked-Vector Method	p. 288
The Extension of Bayesian Classification Theory	p. 290
An Overview	p. 290
Feature Extraction	p. 291
Probability or Evidence Generation	p. 292
Multisource Consensus	p. 292
Bayesian Multisource Classification Mechanism	p. 292
A Refined Multisource Bayesian Model	p. 294
Multisource Classification Using the Markov Random Field	p. 295
Assumption of Intersource Independence	p. 296
Evidential Reasoning	p. 297
Concept Development	p. 297
Belief Function and Belief Interval	p. 299
Evidence Combination	p. 302
Decision Rules for Evidential Reasoning	p. 304
Dealing with Source Reliability	p. 304
Using Classification Accuracy	p. 305
Use of Class Separability	p. 305
Data Information Class Correspondence Matrix	p. 306
The Genetic Algorithm	p. 307
Experimental Results	p. 309
Bibliography	p. 317
Index	p. 349
Table of Contents provided by Ingram. All Rights Reserved.