Road detection using lidar data based on plane assumption and graph model

Detection of traversable road area is one of the key tasks for unmanned platforms to achieve autonomous navigation. We propose a road detection method using lidar data based on plane assumption and graph model in this paper. Two detection algorithms for different speeds are provided to ensure the detection accuracy and to meet the real-time requirements, respectively. First, the point cloud is projected into a real image or an imaginary image for triangulation. The road points are then extended outward from the near platform according to the plane assumption. Then the dual Voronoi diagram of the triangulation is filled according to the attribute of the central element node, and the upper part of the horizon is removed. Finally, image post-processing yields the final detection results. This method perfectly combines the classification of points and the determination of road areas and uses the mature image algorithms in the processing of point cloud data. Experimental tests have been carried out on the public KITTI-Road benchmark, obtaining positive results.