Automatic Building Detection and 2D Footprints Reconstruction for VHR SAR Images

A novel method for building detection and 2D footprint reconstruction for very high resolution synthetic aperture radar was proposed. In the detection part, the generalized gamma distribution (G-Gamma) was firstly used in CFAR (constant false alarm ratio) detection for better modeling both of the homogeneous region and heterogeneous region in SAR image. Because our interests in this chapter focus on the rectangular buildings, some criteria were used to select buildings from all the detection results. In the reconstruction part, optimal rectangular was used to fit the boundaries of the building candidates. Compared with the reference data, experimental results indicate that the proposed method is proved to be with high correct detection ratio and positioning accuracy. For further study, the ground truth data of an interesting area was also measured, and the result shows that the estimated size has a good relationship with the real size of the building.