信息与控制2024，Vol.53Issue(4)：508-519,12.DOI:10.13976/j.cnki.xk.2024.4061
一种用于3D激光雷达点云处理的多目标跟踪算法
A Multi-object Tracking Algorithm for 3D LiDAR Point Cloud Processing
武宏伟 1吕东升 1贾琳2
作者信息
- 1. 北京万集科技股份有限公司,北京 100085
- 2. 中国科学院合肥物质科学研究院,安徽 合肥 230031
- 折叠
摘要
Abstract
As a high-precision sensor,3D LiDAR has improved urban traffic management in recent years.However,urban traffic scenes are complicated,with numerous objects and intersecting traj-ectories.This makes traditional multi-object tracking methods ineffective for accurately generating traffic object motion trajectories.In this study,the real-time tracking of traffic objects in complex urban intersection scenarios is addressed,and a multi-layer object tracking algorithm framework is proposed based on a multi-level structure.The framework combines multi-dimensional feature data association based on elliptical thresholds,adaptive initial value filtering for predicted results,and other methods.Experimental results using roadside perception datasets reveal that the proposed al-gorithm outperforms the initial tracking algorithm.The proposed algorithm enhances the accuracy and robustness of object tracking associations in complex traffic scenes and improves the accuracy of perception tracking.Thus,it has certain engineering application value.关键词
多目标跟踪/轨迹置信度/数据关联/椭圆门限/目标运动模型Key words
multi-object tracking/trajectory confidence/data association/elliptical threshold/target motion mode分类
信息技术与安全科学引用本文复制引用
武宏伟,吕东升,贾琳..一种用于3D激光雷达点云处理的多目标跟踪算法[J].信息与控制,2024,53(4):508-519,12.
