多光线追踪与高度特征约束的激光雷达动态点云检测方法  

LiDAR dynamic point cloud detection method with multi-ray tracing and high feature constraints

作　　者：吕品[1] 过佳珺 赖际舟[1] 方玮 LYU Pin;GUO Jiajun;LAI Jizhou;FANG Wei(Navigation Research Center,College of Automation Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 211106,China)

机构地区：[1]南京航空航天大学自动化学院导航研究中心,南京211106

出　　处：《中国惯性技术学报》2024年第10期1010-1017,共8页Journal of Chinese Inertial Technology

基　　金：国家自然科学基金(62273178);国家自然基金联合基金重点支持项目(U2233215)。

摘　　要：基于激光雷达同时定位与构图方法生成的导航地图通常含有动态点,为保障点云匹配精度及路径规划等任务的正常执行,检测并移除导航地图中的动态点云至关重要。光线追踪算法是一种常用的动态检测方法,然而传统光线追踪算法存在因光线误穿越占据体素导致的体素状态误判问题。针对上述问题,提出了一种多光线追踪与高度特征约束的激光雷达动态点云检测方法。利用邻近射线辅助体素状态的判定,减少假阳性情况;基于高度特征约束方法解决了地面点自干扰问题。室内外数据集上的实验结果表明,所提方法在不同场景下的动态检测性能优于OctoMap方法、Removert方法和ERASOR方法,其中静态保留率高于80%,动态剔除率高于97%。Navigation maps generated based on the simultaneous localization and mapping method(SLAM)of LiDAR usually contain dynamic points,and it is crucial to detect and remove dynamic point clouds in navigation maps in order to guarantee the accuracy of point cloud matching and the normal execution of subsequent tasks such as path planning.Ray tracing algorithm is a commonly used dynamic detection method.However,the traditional ray tracing algorithm has the problem of misjudgment of voxel state due to the ray mistakenly passing through the occupied voxels.To solve the above problems,a LiDAR dynamic point cloud detection method with multi-ray tracing and high feature constraints is proposed.Adjacent rays are used to assist in the determination of voxel states to reduce false positives;and the self-interference problem of ground points is solved based on the height feature constraint method.Experimental results on indoor and outdoor datasets show that the dynamic detection performance of the proposed method is better than that of the OctoMap method,the Removert method and the ERASOR method in different scenarios,where the static retention rate is higher than 80%and the dynamic rejection rate is higher than 97%.

关 键 词：动态检测 激光SLAM 多光线追踪 地面分割 高度特征 

分 类 号：TN958[电子电信—信号与信息处理]