REMOTE SENSING IMAGE ANALYSIS: INCLUDING THE SPATIAL DOMAIN
I BASICS OF REMOTE SENSING I I I INTRODUCTION 125 1 2 HISTORIC OVERVIEW 1 3 CONCEPTS OF REMOTE SENSING 1 3 1 SOURCES OF ELECTROMAGNETIC RADIATION 1 3 2 THE ATMOSPHERE 1 3 3 OBJECT-RADIATION INTERACTION 6 9 9 1 3 4 SENSORS 1 3 5 TRANSMISSION, RECEPTION AND (PRE-)PROCESSING 10 1 3 6 IMAGE ANALYSIS AND INTERPRETATION 10 11 1 3 7 THE FINAL PRODUCT 1 4 THE CHAPTERS IN THIS BOOK 5 5 17 2 SPATIAL VARIABILITY, MAPPING METHODS, IMAGE ANALYSIS AND PIXELS 17 2 1 INTRODUCTION 18 2 2 SPATIAL VARIABILITY IN LANDSCAPES 18 2 2 1 SPATIAL VARIATION 21 HIERARCHICAL MAPPING APPROACHES 22 2 2 3 SPECTRAL PROPERTIES OF VEGETATION 26 2 3 REGISTRATION OF SPATIAL INFORMATION BY REMOTE SENSING SENSORS 29 2 4 STATISTICAL MODELS FOR REMOTE SENSING DATA 33 2 5 THE SPATIAL AND SPECTRAL CLASSIFIER 34 2 6 CONCLUSIONS 37 3 SUB-PIXEL METHODS IN REMOTE SENSING 3 1 INTRODUCTION 37 3 2 MIXED PIXELS 39 3-3 ESTIMATING SUB-PIXEL CLASS COMPOSITION 41 3 3 1 SPECTRAL UNMIXING 41 3-3 2 SOFT CLASSIFICATION 42 3 3 2 1 MAXIMUM LIKELIHOOD CLASSIFICATION 43 3 3 2 2 FUZZY C-MEANS (FCM) 44 3 3 2 3 POSSIBILISTIC C-MEANS (PCM) 3 3 2 4 NEURAL NETWORKS 45 46 3-4 CURRENT RESEARCH TOPICS 47 3-5 CONCLUSIONS 49 4 RESOLUTION MANIPULATION AND SUB-PIXEL MAPPING 51 4 1 INTRODUCTION 51 4 2 SCALE IN REMOTE SENSING 53 4 2 1 DEFINING SCALE 53 4 2 2 SCALES OF MEASUREMENT 54 4 2 2 1 SUPPORT AND PIXEL SIZE 54 4 2 2 2 SPATIAL RESOLUTION AND SPATIAL EXTENT 54 4 2 3 SCALES OF SPATIAL VARIATION 55 4-3 GEOSTATISTICS 56 4 3 1 THE RF MODEL 56 4 3 2 VARIOGRAM ESTIMATION AND MODELLING 4 3 3 INTERPRETING THE MODELLED VARIOGRAM 4 3 4 REGULARIZING THE VARIOGRAM MODEL 57 59 59 4-3 5 KRIGING 60 4-4 CHANGING THE SCALE OF MEASUREMENT 61 4 4 1 UPSCALING 61 4 4 2 DOWNSCALING 63 4 4 2 1 SUB-PIXEL CLASSIFICATION 4 4 2 2 SUPER-RESOLVING A SOFT CLASSIFICATION 4 4 2 3 HOPFIELD NEURAL NETWORK SOLUTION 4 4 2 4 PIXEL-SWAPPING SOLUTION 63 64 65 67 4 4 2 5 CONTINUOUS VARIABLES 4-5 DISCUSSION 4 5 1 APPLICATIONS 4-5 2 ISSUES OF GENERALIZATION AND ACCURACY 4-5-3 ERROR AND THE PSF 4-5-4 FUTURE RESEARCH 4 6 CONCLUSION 5 68 68 68 68 69 70 70 MULTISCALE OBJECT-SPECIFIC ANALYSIS (MOSA): AN INTEGRATIVE APPROACH FOR MULTISCALE LANDSCAPE ANALYSIS 71 5 1 INTRODUCTION 71 5 2 METHODOLOGY 74 5 2 1 REMOTE SENSING DATASET 74 5 2 2 MOSA DESCRIPTION 74 5 2 2 1 OBJECT-SPECIFIC ANALYSIS (OSA) 75 5 2 2 2 OBJECT-SPECIFIC UPSCALING (OSU) 77 5 2 2 3 MARKER-CONTROLLED SEGMENTATION (MCS) 83 5-3 RESULTS 88 5-4 DISCUSSION 5-5 CONCLUSION 90 91 6 VARIOGRAM DERIVED IMAGE TEXTURE FOR CLASSIFYING REMOTELY SENSED IMAGES 93 6 1 INTRODUCTION 93 6 2 TEXTURE AND VARIOGRAM 94 6 2 1 THE VARIOGRAM FUNCTION AS A TEXTURAL DESCRIPTOR 96 6 2 2 GEOSTATISTICAL TEXTURE OPERATORS BASED ON THE VARIOGRAM 99 6 3 CALCULATING GEOSTATISTICAL TEXTURE IMAGES 1?? 103 6 3 1 USING VARIOGRAM-BASED TEXTURE OPERATORS 101 105 6 3 2 USING THE CROSS VALIDATION METHOD 6 4 USING THE VARIOGRAM FUNCTION IN REMOTELY SENSED IMAGE CLASSIFICATION 105 6 4 1 AN EXAMPLE OF USING GEOSTATISTICAL TEXTURE IMAGES IN SUPERVISED DIGITAL CLASSIFICATION 110 6 5 CONCLUSIONS 7 MERGING SPECTRAL AND TEXTURAL INFORMATION FOR CLASSIFYING REMOTELY SENSED IMAGES 113 7 1 INTRODUCTION 113 7-2 BACKGROUND 114 7-3 TEXTURE MEASURES 115 7-4 PER-PIXEL VS PER-FIELD APPROACH TO IMAGE CLASSIFICATION 119 7-5 CLASSIFICATION ALGORITHMS 120 7 6 STUDY AREA 123 7-7 IMAGE CLASSIFICATION 123 7 8 PER-FIELD APPROACH 129 7-9 COMPARING TEXTURE MEASURES 129 133 7 10 COMPARING PER-PIXEL AND PER-FIELD APPROACHES 135 7 11 COMBINING PER-PIXEL AND PER-FIELD APPROACHES 135 7 12 CONCLUSIONS 137 8 CONTEXTUAL IMAGE ANALYSIS METHODS FOR URBAN APPLICATIONS 137 8 1 INTRODUCTION 140 8 2 FREQUENCY-BASED CONTEXTUAL CLASSIFIER 144 8 3 FBC APPLIED TO A MODIFIED FORM OF TEXTURE SPECTRUM 151 8 4 SUMMARY AND CONCLUSIONS 153 9 PIXEL-BASED, STRATIFIED AND CONTEXTUAL ANALYSIS OF HYPERSPECTRAL IMAGERY 153 9 1 INTRODUCTION 155 9 2 PIXEL BASED ANALYSIS OF HYPERSPECTRAL IMAGERY 155 9 2 1 SPECTRAL MATCHING TECHNIQUES USED FOR COMPOSITIONAL MAPPING 160 9 2 2 ABSORPTION BAND PARAMETER ESTIMATES 9-3 STRATIFIED ANALYSIS OF HYPERSPECTRAL IMAGERY 167 9-4 CONTEXTUAL ANALYSIS OF HYPERSPECTRAL IMAGERY 173 9-5 CONCLUSIONS 179 10 VARIABLE MULTIPLE ENDMEMBER SPECTRAL MIXTURE ANALYSIS FOR GEOLOGY APPLICATIONS 181 10 1 INTRODUCTION 181 10 2 THEORY AND METHODS 183 10 2 1 ENDMEMBER SELECTION 183 10 2 2 STANDARDISED UNMIXING 184 10 2 3 VMESMA FEATURES 185 10 3 A CASE STUDY FROM THE AZNALCÓLLAR MINE, SOUTHERN SPAIN 189 10 3 1 INTRODUCTION 189 10 3 2 DATA SET 190 10 3 3 DETECTION OF RESIDUAL CONTAMINATION AND OXIDATION PRODUCTS 190 10 4 A CASE STUDY FROM AKTHARMA-PASHALY MUD VOLCANO, AZERBAIJAN 195 10 4 1 INTRODUCTION 195 10 4 2 ASTER DATA PRE-PROCESSING 197 10 4 3 ASTER MINERAL MAPPING 197 10 4 4 IDENTIFICATION OF KEY ENDMEMBERS 198 10 4 5 RESULTS 199 10 5 CONCLUSIONS 200 II A CONTEXTUAL ALGORITHM FOR DETECTION OF MINERAL ALTERATION HALOS WITH HYPERSPECTRAL REMOTE SENSING 201 11 1 INTRODUCTION 201 11 2 ONSHORE HYDROCARBON SEEPAGES 201 203 11 3 OPTICAL REMOTE SENSING OF ONSHORE HYDROCARBON SEEPAGES 203 11 4 SPECTRAL MODELS OF HYDROCARBON SEEPAGES 204 11 5 COMBINING THE SPECTRAL AND SPATIAL DOMAINS 204 11 5 1 AN ALGORITHM FOR COMBINED SPECTRAL AND SPATIAL DETECTION 207 11 5 2 SIMULATED IMAGERY FOR DEMONSTRATING THE ALGORITHM 208 11 6 RESULTS 210 11 7 DISCUSSION AND CONCLUSIONS 12 IMAGE SEGMENTATION METHODS FOR OBJECT-BASED ANALYSIS AND CLASSIFICATION 12 1 INTRODUCTION THOMAS BLASCHKE, CHARLES BURNETT && ANSSI PEKKARINEN 211 12 1 1 THE RS/GIS IMAGE ANALYSIS CONTINUUM 211 12 1 2 NEW SENSORS, NEW IMAGE/GROUND OBJECT RELATIONSHIPS 212 12 1 3 FROM PIXELS TO IMAGE-OBJECTS 213 12 2 IMAGE SEGMENTATION REVIEW 214 215 12 2 1 WHAT IS IMAGE SEGMENTATION? 215 12 2 2 TYPES OF SEGMENTATION 217 12 2 3 SEGMENTATION OF REMOTELY SENSED DATA: STATE OF THE ART 219 12 2 4 OPERATIONAL IMAGE SEGMENTATION FRAMEWORKS 220 EXTENDING SEGMENTATION TO AN OBJECT-BASED ANALYSIS AND CLASSIFICATION 12 3 223 EXAMPLES OF APPLICATIONS 12 4 225 12 4 1 SEGMENTATION IN MULTI-SOURCE FOREST INVENTORY 225 12 4 2 OBJECT-BASED FOREST STANDS MAPPING IN AN URBAN FOREST 229 12 5 DISCUSSION AND CONCLUSIONS 234 13 MULTISCALE FEATURE EXTRACTION FROM IMAGES USING WAVELETS 237 13 1 INTRODUCTION 237 13 2 MULTIRESOLUTION DECOMPOSITION WITH WAVELETS 238 13 2 1 DIGITAL FILTERS AND FILTER BANKS 241 13 2 2 WAVELETS AND MULTIRESOLUTION 242 13 2 3 ALGORITHMS FOR IMPLEMENTATION 244 13 2 4 OTHER MULTIRESOLUTION DECOMPOSITIONS 247 13-3 RECENT APPLICATIONS IN REMOTE SENSING IMAGE PROCESSING 249 251 13-4 CASE STUDY I - FEATURE EXTRACTION FOR IMAGE REGISTRATION 251 13 4 1 INTRODUCTION 252 13 4 2 IMAGE REGISTRATION BASED ON MULTIRESOLUTION DECOMPOSITION 254 13 4 3 RESULTS AND DISCUSSION 258 13 4 4 CONCLUSIONS 259 13-5 CASE STUDY II - FEATURE EXTRACTION FOR DATA FUSION 259 13 5 1 INTRODUCTION 261 13 5 2 MULTISCALE ANALYSIS FOR IMAGE FUSION 261 13 5 3 TEST SITE AND DATA 262 13 5 4 RESULTS AND DISCUSSION 263 13 5 5 CONCLUSIONS 264 13 6 CASE STUDY III - FEATURE EXTRACTION FOR CHANGE DETECTION 264 13 6 1 INTRODUCTION 265 13 6 2 MULTISCALE PRODUCTS FOR FEATURE EXTRACTION 268 13 6 3 TEST SITE AND DATA 268 13 6 4 RESULTS AND DISCUSSION 269 13 6 5 CONCLUSIONS 271 14 CONTEXTUAL ANALYSES OF REMOTELY SENSED IMAGES FOR THE OPERATIONAL CLASSIFICATION OF LAND COVER IN UNITED KINGDOM 14 1 INTRODUCTION 271 14 2 THE LAND COVER MAP OF GREAT BRITAIN 272 14 2 1 THE SCOPE FOR CONTEXTUAL CORRECTION IN LCMGB 272 14 2 2 LCMGB CONTEXTUAL ANALYSES 273 14 2 3 CONCLUDING COMMENTS ON LCMGB CONTEXTUAL ANALYSES 274 14-3 LAND COVER MAP 2000 275 14 3 1 BACKGROUND 275 14 3 2 BROAD HABITATS AND LCM2000 CLASSES 275 14 3 3 IMAGE SEGMENTATION 277 14-3 4 CONTEXTUAL CLASSIFICATION - MAXIMUM LIKELIHOOD CLASSIFICATION OF PARCELS 279 14-3 5 PARCEL SELECTION FOR POST-CLASSIFICATION CONTEXTUAL ALTERATIONS 280 14 3 6 CONTEXTUAL CORRECTION VERSUS THEMATIC CLASS EXTENSION 282 14-3-7 CONTEXTUAL CORRECTIONS 283 14 3 8 CONTEXTUAL REFINEMENTS AND THEMATIC EXTENSIONS 285 14 3 9 LCM2000 ATTRIBUTE DATA 286 14-3 10 IMPACTS IN QUANTITATIVE TERMS 286 14-4 CONCLUSIONS 289 15 A CONTEXTUAL APPROACH TO CLASSIFY MEDITERRANEAN HETEROGENEOUS VEGETATION USING THE SPATIAL RECLASSIFICATION KERNEL (SPARK) AND DAIS7915 IMAGERY 291 15 1 INTRODUCTION 291 15 2 THE CONCEPT OF SPARK 292 15-3 CASE STUDY 295 15 3 1 INTRODUCTION 295 15 3 2 STUDY AREA 295 15 3 3 VEGETATION 296 15-3-4 DAIS7915 IMAGERY 298 15 3 5 FIELD SURVEY 298 15 3 6 MODIFIED SPARK NDVI APPROACH 299 15 3 7 RESULTS 300 15 3 8 DISCUSSION AND CONCLUSION 309