ABSTRACT
CONTENTS 16.1 Introduction ............................................................................................. 331 16.2 Small-Scale Roof Mapping Using Fused SPOT Pan
and Landsat TM Images........................................................................ 332 16.2.1 Satellite Images ......................................................................... 332 16.2.2 Fusion of Landsat TM and SPOT Pan................................... 333 16.2.3 Multispectral Classification..................................................... 333 16.2.4 Spatial Feature Postclassification ........................................... 335 16.2.5 Accuracy Assessment .............................................................. 337
16.3 Medium-Scale Roof Mapping Using Fused QuickBird Pan and MS Images ............................................................ 339 16.3.1 Satellite Image Used................................................................. 339 16.3.2 Image Fusion ............................................................................. 340 16.3.3 Object-Oriented Classification ................................................ 341
16.3.3.1 Segmentation ............................................................ 342 16.3.3.2 Classification............................................................. 343 16.3.3.3 Results Comparison ................................................ 344
16.4 Conclusions.............................................................................................. 347 References ........................................................................................................... 347
Automatic roof mapping using remote sensing imagery has always been a challenging research topic in remote sensing [1-4]. In general, high spatial resolution and multispectral information are two important information components for automatic classification or extraction of building roofs. The high spatial resolution provides geometric details for delineating
individual objects, whereas the multispectral information supplies color information for differentiating objects with different spectral reflectance.