Right-angle building extraction based on graph-search algorithm

This paper presents a novel graph search schema for right-angle building extraction from high resolution aerial and satellite image subsets which contain a building. As the schema is a edge-driven and bottom-up approach, emphasis is put on elimination of spurious and insignificant low-level image features. In real-world, most buildings are comprised of several sequential corners. We classify the corners into four types according to the orientations of constructing edges. Each type of corners is labeled with a tag to identify the corner, such as ABCD, etc. Based on analysis of right-angle buildings, a set of geometric constraint rules could be derived. Such geometric constraints could then be used in graph construction. Different from tranditional approaches, this proposed approach is implemented by a graph search algorithm to reconstruct building shape.