Extracting road surface from mobile laser scanning point clouds in large scale urban environment

This paper presents a novel automatic road surface extraction method from mobile laser scanning point clouds of large scale complex urban environment. First, the candidate road surface points are automatically separated from the whole point cloud and divided into independent local patches by utilizing the trajectory. Then, each local patch is modeled as a 3D plane and the model parameters are estimated through the generalized projection based M-estimator (GPBM). The points of each local patch are classified into two classes, where points lying within the model are classified as road surface and the others are non-road surface. Finally, a fast clustering method is performed on the road surface class to filter out false positives. The remaining points of all patches constitute the road surface points. Experimental results demonstrate the effectiveness and reliability of the proposed algorithm for automatic road surface extraction in complex urban environments.