Shape-from-Shadow Building Reconstruction from Multiple View SAR Images

Detection and geometric reconstruction of man-made objects from high resolution, multiple view interferometric SAR data gets feasible. IFSAR data is corrupted by blur, speckle noise, and other view dependent effects as e.g., layover and shadows. Especially in case of buildings, these phenomenological features may also provide valuable information about the underlying structure. In this paper we focus on the reconstruction of building walls from multiple view slant range shadows. The exact segmentation of the shadow boundaries is the crucial task of this process. This is accomplished using a rotating mask for edge detection. The mask exploits textural differences between adjacent regions. From the segmented shadow boundaries, the building outlines are reconstructed. The results correspond well to the ground truth data, especially the building’s areas are delimited better than from interferometric data.