Building detection and height estimation from high-resolution insar and optical data

State-of-the-art satellite SAR sensors acquire data of one meter geometric ground resolution, airborne sensors achieve even higher resolutions. Nonetheless, layover and occlusion hamper interpretability of such data particularly in urban scenes. In order to overcome this drawback, we use additional information from aerial photos to detect buildings. Features are extracted from both data sets and introduced to a common feature vector followed by a classification into building sites and non-building sites with Conditional Random Fields (CRF). Furthermore, we show that the different sensor geometries of the SAR and the optical sensor may be used to estimate building heights.