A Minimum-Cost Path Model to the Bridge Extraction from Airborne LiDAR Point Clouds

Nowadays, bridges have played a significant role in human transportation networks. However, less attention has been paid to the bridge extraction from the countryside environment. This paper aims to propose a three-step method for the bridge extraction from airborne LiDAR point clouds. First, we propose a chain-code-based method to delimit land/water interface from the input scene. Second, we perform an angle testing process to extract candidate bridge points based on the shoreline delimitation result. Third, we calculate the cost of paths across the water body. A path whose cost is less than an adaptive threshold will be selected as a bridge path. The main contribution of this paper is that we formulate an energy function to calculate the cost of each potential bridge path. The optimal path, which achieves the minimum cost, is solved by the proposed minimum-cost path model. The developed extraction method does not rely on the geometric shape of rivers and works well in different types of bridges. Experiments show that the presented method succeeds to obtain all bridges in six small bridge scenes and one large complex scene, which are promising results in the bridge extraction.