基于激光雷达点法向量的车辆实时位姿估计  

Real-Time Pose Estimation of Vehicles Based on LiDAR Point Normal Vector

作　　者：张瑞宾 郭应时[1] 陈元华[2] 刘晓刚[2] 周扬[1] Zhang Ruibin;Guo Yingshi;Chen Yuanhua;Liu Xiaogang;Zhou Yang(Chang’an University,Xi’an 710064;Guilin University of Aerospace Technology,Guilin 541004)

机构地区：[1]长安大学,西安710064 [2]桂林航天工业学院,桂林541004

出　　处：《汽车技术》2021年第5期36-40,共5页Automobile Technology

基　　金：国家重点研发计划项目(2019YFB1600500);国家自然科学基金项目(51775053,51908054);广西自然科学基金项目(2020GXNSFAA159071);广西高校中青年教师基础能力提升项目(2019KY0819,2020KY21014);桂林市科学研究与技术开发计划项目(20190206-2);广西职业教育教学改革研究项目(GXGZJG2019A035)。

摘　　要：为解决复杂交通场景下激光雷达获取运动车辆点云残缺而导致位姿信息难以提取的问题,提出一种基于点法向量特征的动态刚性目标实时位姿估计方法。利用最小二乘拟合算法求解目标表面点法向量,采用自适应聚类算法求解点法向量聚簇中心数量,求取聚类点拟合平面,选择互相垂直的平面代表点法向量分别作为待测目标所属坐标系内X轴和Y轴,通过运动学方程完成目标三维姿态角的求解。试验结果表明,该方法能实时估计直线和曲线运动车辆的位姿参数,表现出较好的鲁棒性。In order to solve the problem that it is difficult to extract the pose information of moving vehicles due to the incomplete point cloud acquired by lidar in complex traffic scene,this paper proposes a real-time pose estimation method of dynamic rigid target based on point normal vector feature.The least square fitting algorithm is used to solve the point normal vector of the target surface,and the adaptive clustering algorithm is used to solve the number of clustering centers of the point normal vector.The fitting plane of the clustering points is obtained,and the normal vectors of the representative points in the mutually perpendicular plane are selected as the X-axis and Y-axis in the coordinate system of the target to be measured,and the three-dimensional pose angle of the target is solved through the kinematic equation.The experimental results show that the proposed method can estimate the pose parameters of straight-line and curvilinear vehicles in real time and has good robustness.

关 键 词：激光雷达 运动车辆 点法向量 位姿估计 

分 类 号：U463.67[机械工程—车辆工程]