基于激光雷达的结构化道路障碍物检测方法  

Structured road obstacle detection method based on Lidar

作　　者：王亚波[1] 靳玉良 张亚 范世伟[2] 于飞[2] WANG Yabo;JIN Yuiang;ZHANG Ya;FAN Shiwei;YU Fei(Wuhan Second Ship Design and Research Institute,Wuhan 430205,China;School of Instrument Science and Engineering,Harbin Institute of Technology,Harbin 150001,China)

机构地区：[1]武汉第二船舶设计研究所,武汉430205 [2]哈尔滨工业大学仪器科学与工程学院,哈尔滨150001

出　　处：《中国惯性技术学报》2023年第6期593-600,619,共9页Journal of Chinese Inertial Technology

基　　金：国家自然科学基金(52071121)。

摘　　要：针对传统激光雷达障碍物检测地面分割不充分和固定阈值聚类效果不佳等问题,提出了一种低速行驶状态下改进的结构化道路障碍物检测方法,包括点云数据预处理、点云分割、道路边界提取、点云聚类四个方面。首先获取雷达点云数据并进行预处理;接着将栅格图法和RANSAC算法相结合实现地面点云分割,改善分割效果;然后利用几何特征提取道路边界点,基于邻近栅格信息和边界连续性精细化提取结果;最后引入自适应阈值对DBSCAN聚类算法进行改进,提高聚类准确率。利用KITTI数据集验证算法可行性,实验结果表明所提算法能够实现对于结构化道路边界的准确提取,地面点云分割和道路障碍物检测的准确率均可达到80%以上,与固定阈值相比,效果得到改善。Aiming at the problems of inadequate ground segmentation and poor clustering effect of fixed threshold in traditional LIDAR obstacle detection,an improved structured road obstacle detection method at low speed is proposed,including four aspects:point cloud data preprocessing,point cloud segmentation,road boundary extraction and point cloud clustering.Firstly,the radar point cloud data is acquired and preprocessed.Then,the raster graph method and RANSAC algorithm are combined to realize the ground point cloud segmentation and improve the segmentation effect.Then,the geometric features are used to extract the road boundary points,and the results are refined based on the adjacent raster information and boundary continuity.Finally,adaptive threshold is introduced to improve DBSCAN clustering algorithm to improve clustering accuracy.The KITTI dataset is used to verify the feasibility of the algorithm.Experimental results show that the algorithm can achieve the accurate extraction of structured road boundary,and the accuracy of ground point cloud segmentation and road obstacle detection can reach more than 80%,which is better than the fixed threshold.

关 键 词：激光雷达 点云分割 点云聚类 道路边界提取 障碍物检测 

分 类 号：U666.1[交通运输工程—船舶及航道工程]