Road Boundary Detection Method Based on Region Growing Segmentation Point Cloud

In order to assist intelligent vehicles to correctly identify complex road boundaries, a road boundary detection algorithm based on LiDAR scanning point cloud is proposed. In order to accurately segment the ground point cloud data and avoid the occurrence of over-segmentation and under-segmentation of the point cloud, the region growing method is proposed for fast segmentation. A large amount of ground point cloud data is filtered out of the road area; a multi-threshold road edge feature extraction algorithm is proposed, and multi-parameter thresholds are set by constructing various road edge geometric features to improve the road edge detection accuracy; RANSAC algorithm and distance filtering are used. In order to ensure the accuracy of road edge recognition, Kalman filtering is proposed to predict and track the road edge; by collecting real scene point cloud data to verify the algorithm. The results show that the algorithm can effectively identify the road edge, the ground segmentation algorithm has a high accuracy and can avoid over-segmentation, which verifies the robustness and accuracy of the algorithm.