Curb Detection Method Based on Curvatures for Urban Autonomous Navigation

Abstract—This paper addresses the problem of curb detec-tion for ADAS or autonomous navigation in urban scenarios. The algorithm is based on clouds of 3D points and it is evaluated using 3D information from a pair of stereo cameras and also from a LIDAR. The feature to detect the curbs is the curvature of the road surface. The curvature estimation requires a dense point cloud, therefore the density of the LIDAR cloud has been increased using Iterative Closest Point (ICP) with the previous scans. The proposed algorithm can deal with curbs of different curvature and heights, from as low as 3 cms, in a range up to 20 m if the curbs are connected in the curvature image. The parameters of the curb are modeled with lines and lateral error is compared to a ground truth. The ground truth sequences were manually labeled by an expert and they are publicly available for the community