Detection and Reconstruction of Building Footprints from Single VHR SAR Images

Abstract:
Recently, the development of automatic methods for building detection and reconstruction from new generation spaceborne very high resolution (VHR) synthetic aperture radar (SAR) images has become a challenging topic in many application scenarios (e.g. damage assessment). In this paper we present a novel method for automatic building detection which also reconstructs the 2D footprint of the detected buildings. Simple features are extracted from the SAR image and then composed to building candidates exploiting their semantic meaning. The 2D reconstruction aims at correcting errors arising in the feature extraction step and at estimating the size and aspect angle of the buildings.