Abstract
Lane line recognition and classification is an important content in the field of advanced assisted driving and autonomous driving. Based on LiDAR (Light Detection and Ranging), this paper proposes a fine lane recognition method, namely dashed and solid lane lines classification, and verifies its feasibility. First, the collected LiDAR data is preprocessed, then the lane points in the point cloud are detected, and finally the lane line fitting is carried out for the detected lane points. In the preprocessed process, KD-Tree is used to search the point cloud in the ROI (Region of interest), and then RANSAC (Random Sample Consensus) algorithm is used to fit it. An appropriate threshold is set to extract the optimal plane to get the ground data, and then the window search is used to iteratively search lane points one by one along the road direction, and the final lane lines are obtained by parabolic fitting and parallel fitting of lane points. Dashed and solid lines can be distinguished according to the mean reflection intensity of point clouds on the lane line. Based on MATLAB, the method is verified, and the results show that the method can accurately identify the lane line, and can classify the dashed lines and solid lines.