SAR Data Classification of Urban Areas by Means of Segmentation Techniques and Ancillary Optical Data

The classification of urban areas in terms of land-use/land-cover (LULC) maps is a challenging as well as essential task in order to monitor how the urban sprawl is changing the environment. This paper is devoted to the description of a novel procedure designed to exploit coarse-resolution SAR images and obtain both the built-up area extents and a LULC map of the individuated urban area. The approach starts from the previously developed BuiltArea algorithm to produce the built-up area extent map, exploiting the spatial correlation among neighboring pixels by means of local indicators of spatial association and gray level co-occurrence matrix (GLCM) features. After discriminating between urban and nonurban areas, a novel approach is presented that exploits segmentation techniques, spatial feature selection, and a supervised classifier to generate urban LULC maps. A robust chain, considering SAR data and using ancillary optical data is proposed and validated using data sets available in two test cases, the megacities of Shanghai and Beijing.