基于目标检测的激光雷达道路边缘检测算法

LiDAR Road Edge Detection Algorithm Based on Target Detection

道路边缘检测是自动驾驶车辆环境感知的重要组成部分,有效地从点云数据中提取道路边缘信息,有利于进行目标检测以及可行驶区域检测。针对点云道路边缘检测问题,提出了一种考虑车辆等道路参与者对道路边缘检测带来干扰的解决方案。首先,采用地面点云分割算法,将原始点云分割成地面点云和非地面点云;其次,根据车辆等道路参与者的固有特性,采用点云聚类算法对点云进行聚类,并将符合车辆等道路参与者特性的非地面点云进行滤除;再次,根据道路边缘点云在二维平面内,能够有效地遮挡激光发射中心点与非道路边缘点之间的连线,从而提取道路边缘点云;最后,采用随机抽样一致性(random sample consensus,RANSAC)算法对道路边缘点云进行多项式拟合,并使用扩展卡尔曼滤波器对道路边缘进行跟踪。实验结果表明,所提点云道路边缘检测算法能够消除车辆等道路参与则对点云道路边缘检测的影响,且算法满足实车实时性和鲁棒性要求。

Road edge detection is an important part of the environment perception of autonomous vehicles.Effectively extraction road edge information from point cloud data is beneficial for target detection and drivable area detection.A solution was proposed to address the issue of point cloud road edge detection,taking into account the interference of road participants such as vehicles in road edge detection.Firstly,a ground point cloud segmentation algorithm was used to divide the original point cloud into ground and non-ground points.Secondly,based on the intrinsic characteristics of vehicles and other road participants,a point cloud clustering algorithm was employed to cluster the points and filter out cloud non-ground points that meet the characteristics of vehicles and road participants.Thirdly,considering that road edge point clouds can effectively occlude the line connecting the laser emission center point with non-road edge points in the two-dimensional plane,the road edge point cloud was extracted.Finally,the random sample consensus(RANSAC)algorithm was used to perform polynomial fitting on the road edge point cloud,and an extended Kalman filter was utilized to track the road edge.Experimental results demonstrate that the proposed point cloud road edge detection algorithm can eliminate the interference of vehicles and other road participants on road edge detection,while satisfying real-time and robustness requirements for practical vehicle applications.