Models for Features Extraction from Sar Images of Urban Areas

Features extraction from SAR images of urban scenes represents one of the most interesting subjects of research in the field of remote sensing; new, more adequate and specialised backscattering and radar models for features retrieval are studied and strongly required. In this paper, a synoptic view of three different approaches to the problem, accounting for the dynamism and complexity of urban scenarios, is presented. Stochastic analysis and neural networks based approaches are here analysed and compared together with a new theory for a deterministic extraction (inverse problem) of geometric and electromagnetic building parameters from the corresponding SAR image.