Building change detection in satellite stereo imagery based on belief functions

3D Building change detection has become a popular research topic along with the improvement of image quality and computer science. When only building changes are of interest, both the multi-temporal images and Digital Surface Models provide valuable but not comprehensive information in the change detection procedure. Therefore, in this paper, belief functions have been adopted for fusing information from these two sources. In the first step, two change indicators are proposed by focusing on building changes. Both indicators have been projected to a sigmoid curve, in which both the concordance and discordance indexes are considered. In order to fuse the concordance and discordance indexes and further fuse the two change indicators, two belief functions are considered. One is the original Dempster-Shafer Theory (DST), and the most recent one is Dezert-Smarandache Theory (DSmT). This paper shows how these belief-based frameworks can help in building change detection problem. Besides using different belief functions in obtaining the global BBAs, four decision-making criteria are tested to extract final building change masks. The results have been validated by compared to the manually extracted change reference mask.