Detection and classification of urban structures based on high-resolution SAR imagery

In this paper, several advanced methods for detecting and classifying urban structures in high-resolution SAR imagery are presented. The detection of man-made structures resp. the suppression of clutter, for example natural surfaces, can be accomplished by a coherent time-frequency analysis. These techniques can detect targets in the scene that behave like point-scatterers (such as edges, poles, etc.) by their high degree of correlation across sub-apertures of the scene. If polarimetric data are available, this approach can be extended by taking into account the frequently anisotropic polarimetric backscattering of man-made targets. As a third approach, a simple but powerful operator for detecting urban structures by analysing texture inhomogeneity in SAR images is developed. A second objective of this paper is the segmentation and classification of the detected urban structures. Based on the detected man-made structures, the characteristic structure size is investigated, using several approaches, to distinguishing between residential areas, industrial areas and dense city centres. The methods are evaluated using several real airborne SAR datasets.