一种多线激光雷达与GNSS/INS系统标定方法

对于车载激光雷达(LiDAR),惯导系统(INS)与卫星导航(GNSS)的组合导航定位系统,传感器数据融合之前需要对传感器安装位置进行标定。针对以上目标本文提出了一种基于分步迭代法与KD树优化的DBSCAN算法寻求最优安置参数的方法。对初始安置参数进行粗略测量后通过分步迭代法迭代初始参数附近的值,并基于KD树优化的密度聚类算法(DBSCAN)将转换到唯一坐标系下的相邻帧点云集位置重合度进行评价,择优得到最准确的安置参数。实验使用最近迭代点算法(ICP)对不同分步步长的标定参数结果进行对比,实验结果表明安置参数在缩小步长的情况下精度会得到略微提高;本文采用的方法相比初始估计外参提高了标定精度。

A calibration method of lidar and GNSS/INS system

For Light Detection And Ranging(LiDAR),Inertial Navigation(INS) and Global Navigation satellites system(GNSS) integrated navigation and positioning systems,the sensor mounting location needs to be calibrated before the sensor data is fused.Aiming at the above objectives,this paper proposes a method based on fragmented iterative method and KD tree optimization DBSCAN algorithm to find the optimal placement parameters.After the initial measurement parameters are roughly measured,the values near the initial parameters are iterated by the fragmented iterative method,and the DBSCAN algorithm based on the KD tree optimization is used to evaluate the coincidence degree of the adjacent frame points in the unique coordinate system,and the best accurate placement parameters is obtained.The experiment uses Iterative Closest Point(ICP) algorithm to compare the calibration parameters of different stepsizes.The experimental results show that the accuracy of the placement parameters is slightly improved in the case of reducing the step size.The method used in this paper improves the calibration accuracy compared with the initial estimation of external parameters.