利用平面特征的激光雷达相对位姿标定方法

Relative Pose Calibration Method of Lidar Based on Plane Features

针对多激光雷达外参数的标定问题,提出一种利用平面匹配的激光雷达间相对位姿自动标定方法。首先利用RANSAC算法对回廊式结构进行平面和相应法向量的提取;其次通过一系列的方法匹配对应平面,采用Kabsch算法求解初始外参数;最后利用Ceres库进行非线性优化,实现了较高精度的激光雷达外参数标定。仿真实验以及实际测试结果表明,该方法具有较高的标定精度和较强的鲁棒性,旋转、平移误差分别小于0.03 rad和0.04 m。

Aimed at the external parameters calibration of multi lidars, an automatic calibration method of relative position and pose between lidars based on plane matching is proposed in the paper. Firstly, RANSAC algorithm is used to extract the plane and corresponding normal vectors of the corridor structure. Secondly, a series of procedures are used to match the corresponding planes, and then Kabsch algorithm is used to solve the initial external parameters. Finally, Ceres library is used for nonlinear optimization to achieve more accurate calibration. Simulation experiments and practical tests show the method has high calibration accuracy and strong robustness, and the rotation and translation errors are less than 0.03 rad and 0.04 m respectively.