Automatic reconstruction of 3-d building model from airborne tomosar point clouds

This paper proposes an enhanced framework for reconstructing three-dimensional (3-D) building model from airborne tomographic synthetic aperture radar (TomoSAR) point clouds which involves four crucial steps: progressive facade detection, extraction of roof points, extraction of roof outlines and reconstruction. Firstly, an efficient and robust building facade detection is undertaken by adopting a progressive detection method. Then, to extract roof points, the region growing procedure is improved by utilizing an adaptive neighborhood and robust normals of points. Subsequently, the ff-shape algorithm is executed and enhanced through Delaunay triangulation network to extract fine outline points. Finally, roof outlines are regularized, and the building models get reconstructed. The proposed approach is tested using airborne To-moSAR point clouds in the Emei area located in the Sichuan province of China. The results show that the proposed method can reconstruct 3-D building models with better shapes compared to classical methods.