Abstract:

Buildings are an important target of the monitoring of geographical conditions. The rapid and accurate extraction of urban buildings can bring great economic value. Many people have done a lot of work in the building extraction of the city area. On the basis of predecessors' research and aiming at the problems of existing extraction method, this paper proposes an object-based automatic building extraction process in dense urban areas. First, high-resolution remote sensing images are used to get the shadow and quasi building extraction results. Then, build a filter by the spatial location relationship of the shadows and the crude building extraction results to filter the suspected building area. Finally, get a precise building outline through the graph cut algorithm. For algorithm verification experiments, the accurate detection results can be obtained by using two QuickBird images in Wuhan. This algorithm can be applied to the building detections in dense urban areas.

Key words: high-resolution satellite image, object-based, building extraction