Page 13 - Digital Analysis of Remotely Sensed Imagery

xii Co n t e n t s

7.3.3 Agglomerative Hierarchical
Clustering . . . . . . . . . . . . . . . . . . . . . . . 264
7.3.4 Histogram-Based Clustering . . . . . . . 266
7.4 Supervised Classification . . . . . . . . . . . . . . . . . . . 267

7.4.1 Procedure . . . . . . . . . . . . . . . . . . . . . . . . 267
7.4.2 Selection of Training Samples . . . . . . . 270
7.4.3 Assessment of Training
Sample Quality . . . . . . . . . . . . . . . . . . . 271

7.5 Per-Pixel Image Classifiers . . . . . . . . . . . . . . . . . . 271

7.5.1 Parallelepiped Classifier . . . . . . . . . . . 272
7.5.2 Minimum-Distance-to-Mean Classifier . . . 274

7.5.3 Maximum Likelihood Classifier . . . . . 276
7.5.4 Which Classifier to Use? . . . . . . . . . . . 281

7.6 Unsupervised and Supervised Classification . . . 283

7.7 Fuzzy Image Classification . . . . . . . . . . . . . . . . . 284
7.7.1 Fuzzy Logic . . . . . . . . . . . . . . . . . . . . . . 285
7.7.2 Fuzziness in Image Classification . . . 287

7.7.3 Implementation and Accuracy . . . . . . 289

7.8 Subpixel Image Classification . . . . . . . . . . . . . . . 291
7.8.1 Mathematical Underpinning . . . . . . . 291
7.8.2 Factors Affecting Performance . . . . . . 293
7.8.3 Implementation Environments. . . . . . 294
7.8.4 Results Validation . . . . . . . . . . . . . . . . . 296
7.9 Postclassification Filtering . . . . . . . . . . . . . . . . . . 297

7.10 Presentation of Classification Results . . . . . . . . . 300

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
8 Neural Network Image Analysis . . . . . . . . . . . . . . . . . 305
8.1 Fundamentals of Neural Networks . . . . . . . . . . 306
8.1.1 Human Neurons . . . . . . . . . . . . . . . . . . 306

8.1.2 Artificial Neurons . . . . . . . . . . . . . . . . . 306
8.2 Neural Network Architecture . . . . . . . . . . . . . . . 307
8.2.1 Feed-Forward Model . . . . . . . . . . . . . . 309
8.2.2 Backpropagation Networks . . . . . . . . 311
8.2.3 Self-Organizing Topological Map . . . 313
8.2.4 ART . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
8.2.5 Parallel Consensual Network . . . . . . . 316
8.2.6 Binary Diamond Network . . . . . . . . . . 317
8.2.7 Structured Neural Network . . . . . . . . 317
8.2.8 Alternative Models . . . . . . . . . . . . . . . . 319
8.3 Network Learning . . . . . . . . . . . . . . . . . . . . . . . . . 321
8.3.1 Learning Paradigm . . . . . . . . . . . . . . . . 321
8.3.2 Learning Rate . . . . . . . . . . . . . . . . . . . . 322
8.3.3 Learning Algorithms . . . . . . . . . . . . . . 323
8.3.4 Transfer Functions . . . . . . . . . . . . . . . . 324