基于动态补偿与残差估计的惯导和雷达标定方法  

Inertial Guidance and Radar Calibration Method Based on Dynamic Compensation and Residual Estimation

作　　者：胡盛辉 梁冬泰[1] 孙浩铭 夏金泽 王泽华 Hu Shenghui;Liang Dongtai;Sun Haoming

机构地区：[1]宁波大学机械工程与力学学院,浙江宁波315211 [2]诠航科技有限公司,浙江宁波315100

出　　处：《机械制造》2023年第5期68-71,76,共5页Machinery

基　　金：浙江省公益技术应用研究计划项目(编号:LGG21E050008);宁波市交通运输科技项目(编号:202105);宁波市鄞州区农业与社会发展科技项目(编号:20211YZQ010109)。

摘　　要：针对惯导和雷达联合标定方法在小场景下标定精度低的问题,提出一种基于动态补偿与残差估计的惯导和雷达标定方法。这一标定方法对获得的激光雷达点云进行分割,通过点云特征分类,实现点云坏点剔除。利用动态补偿方法,对有效特征点去畸变,获得激光雷达实时位姿信息。对惯导和激光雷达的自身位姿进行外参标定,并对得到的初始变换矩阵进行时间戳对齐。利用几何位姿特征构建残差估计,由此得到最优的位姿变换和地图。实际环境验证结果表明,采用这一标定方法得到的点云地图无重影,输出的位姿轨迹平缓稳定,能够达到预期的标定效果。Aiming at the problem of low calibration accuracy of inertial guidance and radar joint calibration method in small scenario,an inertial guidance and radar calibration method based on dynamic compensation and residual estimation was proposed.This calibration method segments the obtained lidar point cloud,and realizes the point cloud dead pixel removal through the point cloud feature classification.The dynamic compensation method is used to distort the effective feature point and obtain real-time pose information of lidar.The external parameters of the inertial guidance and lidar postures were calibrated,and the obtained initial transformation matrix is time stamped and aligned.The geometric pose feature is used to construct residual estimation,and the optimal pose transformation and map are obtained.The actual environment verification result shows that the point cloud map obtained by this calibration method has no ghosting,and the output pose trajectory is smooth and stable,which can achieve the expected calibration effect.

关 键 词：动态补偿 残差估计 惯导 雷达 标定 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]