ABSTRACT
CONTENTS 8.1 Introduction ............................................................................................... 143 8.2 Impervious Area Mapping in Challenging Environments................ 145 8.3 Methodology.............................................................................................. 146
8.3.1 Overall Procedure......................................................................... 146 8.3.2 Fusion Rule Base........................................................................... 147 8.3.3 Spatial Analysis with Multiple-Width Textures...................... 147 8.3.4 Markov Random Field Classifier ............................................... 149
8.4 Experimental Results................................................................................ 150 8.4.1 Al-Fashir (Sudan).......................................................................... 151 8.4.2 Pavia (Italy).................................................................................... 153
8.5 Conclusions................................................................................................ 156 Acknowledgments ............................................................................................. 157 References ........................................................................................................... 158
Urban environment is by far the most complex one that may possibly appear in remotely sensed images, and its analysis requires extracting a wealth of information from the sensed data. On the one hand, identification of very different land-cover classes is required. On the other hand, spatial patterns should be considered to associate land-cover classes to land-use classes and to discriminate between natural and artificial objects. As a result, any single sensor may contribute to urban remote sensing, but no one is in itself sufficient to capture all the available information. In this chapter we therefore deal with data fusion, that is, the idea of using datasets coming from different sensors, to identify impervious surfaces. To limit the scope of our discussion, we focus on synthetic aperture radar (SAR) and optical data,
which is at the same time our field of experience in the area and one of the most relevant cases, because of, if nothing else, the sheer quantity of data available.
