摘要
为了保证露天矿山运输车辆的运行安全,提高非平铺道路、扬尘环境下障碍物检测与跟踪的高效性和鲁棒性,文章提出一种基于激光雷达与毫米波雷达融合的多目标检测与跟踪算法。该算法包括点云特征提取模块、目标跟踪模块以及异构传感器融合模块。其中,特征提取模块采用点云梯度与距离特征提取方法,解决了多变场景下的小型和异形障碍物点云分割问题;目标跟踪模块构造航迹信息进行多目标跟踪,提高了目标跟踪的稳定性;异构传感器融合模块被设计了多源传感器异步融合策略,克服异构传感器融合难题,提升了矿区非铺装道路和扬尘环境下各异目标检测与跟踪能力。最后,基于机器人操作系统ROS框架,将该算法在露天矿山运输车辆上进行了实车测试。结果显示,系统不仅能在200 m范围内精确检测到矿卡,而且能精确检测到50 m范围内30 cm见方的目标物,表明采用所提算法的系统能够稳定可靠地检测目标并实现跟踪。
To ensure the safety of mine trucks on unpaved roads and in dusty environments, it's necessary to improve the efficiency and robustness of targets detection and tracking. This paper proposes a multi-target detection and tracking algorithm based on the fusion of light detection and ranging (LiDAR) and millimeter-wave radar. The algorithm includes point cloud feature extraction module, target tracking module and heterogeneous sensor fusion module. Among them, feature extraction module uses point cloud gradient and distance feature extraction methods to solve the point cloud segmentation problem of small and irregular-shaped obstacles in changing scenes;target tracking module constructs track information for multi-target tracking, which improves the stability of target tracking;an asynchronous fusion strategy of multi-source sensors is designed for heterogeneous sensor fusion module, which overcomes the problem of heterogeneous sensor fusion, and improves the detection and tracking capabilities of small and various targets in unpaved road and dusty environment of open-pit mines. Finally, verification experiment is carried out on the mine trucks based on ROS(robot operating system) framework and the results show that it can accurately detect mine trucks in the range of 200 m and objects of 30 cm in the range of 50 m, which demonstrate the stability, accuracy and reliability of the method.
作者
李源征宇
胡云卿
龙腾
黄文宇
潘文波
李培杰
LI Yuanzhengyu;HU Yunqing;LONG Teng;HUANG Wenyu;PAN Wenbo;LI Peijie(CRRC Zhuzhou Institute Co.,Ltd.,Zhuzhou,Hunan 412001,China)
出处
《控制与信息技术》
2022年第5期80-85,共6页
CONTROL AND INFORMATION TECHNOLOGY
基金
国家重点研发计划(2021YFB2501800)。
关键词
露天矿山
点云分割
多源传感器融合
环境感知
目标检测
open pit
point cloud segmentation
multi-source information fusion
environmental perception
target detection
