A sparse reconstruction and fusion technique for building footprint extraction from multi-aspect high-resolution SAR data

Traditional synthetic aperture radar (SAR) image processing for building detection and reconstruction faces the problems of sidelobe interference, geometric distortions and surrounding environment aliasing. In this paper, a sparse reconstruction and fusion technique is proposed for building footprint extraction from multi-aspect high-resolution SAR data. It can process different kinds of buildings including flat-roof buildings and gable-roof buildings. The building wall-ground double-bounce is mainly taken into account for its strong scattering properties and sparsity. First, double-bounce is extracted by compressive sensing (CS)-based sparse reconstruction and Hough transform. Then, multi-aspect data are projected and fused, and L-structures are extracted afterward. Finally, qualification by scores of confidence is proposed for the test of footprint candidates, and building footprint map is derived. Numerical experiment and comparison with traditional methods are carried out to show the effectiveness and performance improvement of the proposed method, with far fewer echo samples used.