Extraction of Bridges from High Resolution Remote Sensing Image Based on Topology Modeling

Bridges are the hubs of transportation, so it is important to identify and locate bridges for satellite image interpretation. This paper proposes a new method of bridge extraction from high resolution remote sensing images. Firstly, river regions are segmented and road lineaments are extracted. Then, bridge region is represented as intersection of river with roads or roads with roads by using the recognition model proposed in this paper. Finally, a rule-based procedure is applied to verify candidate regions. The experiment results show that, not only bridges over river but also the overpasses can be extracted effectively. The method mainly uses structural information of lineaments on roads and the topological relations in bridge regions, and does not rely on accurate results of river segmentation, so it is robust for complex scenes.