基于点线面特征匹配的紧耦合激光雷达惯性里程计  

Tightly coupled LiDAR inertial odometry based on point-line-surface feature matching

作　　者：刘士良 马天力[1] 高嵩[1] 严瀚宇 LIU Shiliang;MA Tianli;GAO Song;YAN Hanyu(School of Electronic Information Engineering,Xi’an Technological University,Xi’an 710021,China)

机构地区：[1]西安工业大学电子信息工程学院,陕西西安710021

出　　处：《传感器与微系统》2024年第7期72-76,共5页Transducer and Microsystem Technologies

基　　金：陕西省重点研发计划资助项目(2022GY—242);陕西省技术创新引导专项(基金)计划资助项目(2022QFY01—16)。

摘　　要：激光雷达(LiDAR)里程计在室外受环境噪声干扰,其扫描匹配精度较低,由此带来的累积误差导致同步定位与建图(SLAM)在大范围场景下定位精度较差。针对上述问题,提出一种基于点线面特征匹配的紧耦合LiDAR惯性里程计(TP-LIO),在惯性测量单元(IMU)预积分点云去畸变基础上,利用点—线和点—面距离构建位姿估计代价函数,获得载体运动估计。结合IMU预积分与Scan-Context回环检测,通过因子图进行全局优化,实现IMU和LiDAR数据紧耦合。在KITTI数据集和实车实验下,对TP-LIO、ALOAM和LeGO-LOAM进行对比验证,结果表明:TP-LIO在大范围场景下累积误差更小,定位精度更高。LiDAR odometry is disturbed by environmental noise outdoors,which cause low scan matching precision,accumulated error caused by scan matching leads to poor positioning precision of simultaneous localization and mapping(SLAM)in large-scale scenes.Aiming at the above problem,a tightly-coupled LiDAR inertial odometry(TP-LIO)based on point-line-surface feature matching is proposed.On the basis of inertial measurement unit(IMU)pre-integration point cloud de-distortion,point-line and point-surface distances are used to construct pose estimation cost functions to obtain carrier motion estimation.Combined with IMU pre-integration and Scan-Context loop detection,global optimization is performed through factor graphs to achieve tight coupling of IMU and LiDAR data.In the KITTI dataset and real vehicle experiments,TP-LIO,ALOAM and LeGO-LOAM are compared and verified.The results show that TP-LIO has smaller cumulative error and higher positioning precision in a wide range of scenes.

关 键 词：同步定位与建图 激光雷达 紧耦合 回环检测 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]