基于点特征的车载LiDAR与GNSS/IMU联合标定方法研究

激光雷达(LiDAR)、全球导航卫星系统(GNSS)与惯性测量单元(IMU)之间的外部参数标定精度是影响多传感器融合及高精度地图的主要因素。基于此,提出一种适用于无人车的LiDAR与GNSS/IMU标定方法,该方法可实时提取标定板点云中心坐标,并利用点特征进行外部参数标定。首先,分析了LiDAR、GNSS、IMU及通用横墨卡托格网系(UTM)坐标系的变换关系;其次,基于IMU安装平面与地面平行的假设,通过车辆前方地面的法向量计算LiDAR俯仰角和滚转角的初值,并利用标定板中心偏移量计算偏航角初值;然后,假设车辆在平面上保持直线运动,采用恒定姿态运动将旋转角度的求解问题转化为最优化问题,并根据标定板UTM坐标不变的约束求解出平移参数;最后,通过分析算法误差、传感器测量误差以及中心点匹配误差验证了方案的可行性,并通过无人矿车采集LiDAR、GNSS和IMU的同步数据,对所提出的外部参数标定方法进行测试。试验结果表明：提出的方法只需要一定范围的平坦区域和标定板就可以解决3自由度运动估计6自由度外部参数的退化问题,且标定后的外部参数可以保证点云地图在20 m内的拼接误差小于20 cm。

Research on Joint Calibration Method of LiDAR and GNSS/IMU Based on Point Feature

The accuracies of extrinsic parameter calibration among the light detection and ranging (LiDAR), global navigation satellite systems (GNSSs), and inertial measurement unit (IMU) are the main factors affecting multisensor fusion and high-definition maps. This study presents a LiDAR and GNSS/IMU calibration method for unmanned vehicles that extracts the point cloud of a calibration pattern in real time and uses point features to calibrate extrinsic parameters. First, the coordinate transformation between LiDAR, GNSS, IMU, and Universal Transverse Mercator Grid System (UTM) was analyzed. Second, assuming that the installation plane of the IMU is parallel to the ground, the initial values of pitch and roll were calculated based on the normal vector of the ground in front of the vehicle, and the initial value of yaw was calculated using the offset of the calibration pattern center. Third, based on the assumption that the vehicle moves straight on a plane, constant attitude motion was used to transform the problem of solving the rotation angle into an optimization problem by solving the translation parameters according to the constraint of the unchanged UTM coordinate of the calibration pattern. The feasibility of the scheme is verified by analyzing the error of the algorithm, sensor measurement, and center-point matching. An unmanned mining truck was used to collect LiDAR, GNSS, and IMU synchronization data to test the proposed method. The experimental results demonstrate that the proposed method requires only a flat area and calibration pattern to solve the degradation problem of 6-DOFs extrinsic parameters resulting from 3-DOFs motion, and the calibrated extrinsic parameters ensure that the splicing error of a point cloud map within 20 m is less than 20 cm.