摘要
本文提出了一种低速运动情况下使用稀疏点云的实时激光里程计方法。利用局部曲率和主成分分析提取局部网格内的点云面特征,使用点-面距离作为匹配代价函数,在点云位姿估计中利用点云距离和特征平面性对数据重要性进行加权,并针对低速运动对位姿变换矩阵进行近似。在包含大量不稳定特征的场景下,使用16线旋转式机械激光雷达采集的稀疏点云序列进行了试验。结果表明,本文方法对稀疏激光雷达点云具有很好的稳定性,针对低速运动所做近似是有效的,能够大幅提高计算的实时性。
A novel algorithm for low-latency LiDAR odometry is proposed for low-speed movement with sparse point cloud.The plane features was extracted based on curvature using principal component analysis.The distance between point and the nearest plane was adopted as the cost in the matching.For the low-speed movement,the transform matrix between two point clouds was simplified as an linear least-square problem and the data was weighted by the distance of the cloud point and the planarity of the local plane.The experiments were conducted in the scenarios with unstable features using spin LiDAR with 16 beams.The results shows the stability for the sparse point cloud of the proposed method.The approximation of the transform matrix for the low-speed movement is effective and the latency is low.
作者
冯宝新
FENG Baoxin(Tongji University,Shanghai 200092,China)
出处
《测绘通报》
CSCD
北大核心
2024年第S01期137-142,共6页
Bulletin of Surveying and Mapping
关键词
激光里程计
稀疏点云
实时匹配
低速运动
LiDAR odometry
sparse cloud point
real-time matching
low-speed movement
作者简介
冯宝新(1974—),男,博士,高级工程师,主要研究方向为低线束激光雷达在测绘中的应用。E-mail:522406642@qq.com
