基于3D激光雷达点云的道路边界识别算法  

Road boundary identification algorithm based on 3D LIDAR point cloud

作　　者：孔栋[1] 孙亮[1] 王建强[2] 王晓原[1,2] 

机构地区：[1]山东理工大学交通与车辆工程学院,山东淄博255049 [2]清华大学汽车安全与节能国家重点实验室,北京100084

出　　处：《广西大学学报（自然科学版）》2017年第3期855-863,共9页Journal of Guangxi University（Natural Science Edition）

基　　金：国家自然科学基金资助项目(61074140;61573009;51508315;51608313);汽车安全与节能国家重点实验室开放基金(KF16232);山东省自然科学基金(ZR2014FM027;ZR2016EL19);山东省社会科学规划研究项目(14CGLJ27);山东省高等学校科技计划(J15LB07)

摘　　要：针对结构化道路环境中道路边界存在不连续、被遮挡及易受路内障碍物干扰情况下的识别问题,利用车载激光雷达获取的结构化道路环境三维点云数据的高程信息,结合局部均值变点统计方法,提出了一种用于激光雷达帧数据的道路边界识别算法。该算法首先利用局部均值变点统计对结构化道路环境三维点数据中突变的z坐标值进行标记并提取其对应的(x,y)数据点,即道路边界点数据粗提取;然后基于分段双阈值对粗提取的道路边界点数据滤波处理;最后利用最小二乘法拟合道路边界点数据。基于实车实验分别采集的不同道路环境条件下结构化直道1 450帧、弯道935帧数据,算法识别准确率均高于80%,且识别道路宽度误差小于0.14 m。实验结果表明,该算法不仅能够自动识别结构化道路边界,而且有效抑制了路面障碍物的干扰,验证了算法的有效性。Aiming at the road boundary identification for discontinuous,occluded and susceptible to obstructions in the road structured road environment, a road boundary identification algorithm for LIDAR （ Light Deteation and Ranging） data frames is proposed by using the local mean change point statistics based on the elevation information of structured road environment points acquired by the onboard lidar. Firstly,the local mean change point statistics is used to mark the z-coordinate of the mutation in the data of the structured road environment and to extract the corresponding （x , y） data points,rough extraction of the road boundary point data. Then,the segmented dual threshold method is used to filter the road boundary points. Finally, the least squares method is used to fit the road boundary point data. Based on the laser radar acquisition of structured straight road 1450 frame data and bending road 935 data under different road environment conditions, the accuracy of the algorithm is higher than 80% , and the recognition error of the road width is less than 0.14 m . The experimental results show that the algorithm can not only automatically identify the structured road boundary, but also effectively restrain the obstacle interference, and verify the effectiveness of the algorithm.

关 键 词：智能车 激光雷达 均值变点统计 道路边界识别 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]