3D Reconstruction of Buildings and Vegetation from Synthetic Aperture Radar (SAR) Images

D models of a landscape are useful for the eval- uation of remote sensing data or for landscape plan- ning. In this paper, it is shown that a 3D reconstruc- tion of buildings and vegetation can be obtained from the shadows in synthetic aperture radar (SAR) images. First, elevated objects like buildings and higher vegetation (e. g. trees), and shadow regions are segmented in the SAR image. Then the shad- ows are assigned to suitable elevated objects. Since it is assumed that the height of the ground is known from a digital elevation model (DEM) or from other sensors, the heights of the elevated objects can be calculated from the estimated radial shadow lengths and the system parameters. The wrong positions of elevated objects in SAR images can be corrected af- terwards. Finally, triangulation and texturing meth- ods lead to a realistic 3D model of the landscape including buildings and vegetation.