基于多传感器的建筑工业化吊运装备运行过程障碍物检测  

Multi-sensor Fusion-based Obstacle Detection During the Operation of Building Industrialized Lifting Equipment

作　　者：凌鹤 位权权[2] 潘基铎 卢红 LING He;WEI Quan-quan;PAN Ji-duo;LU Hong(Hubei Provincial Key Laboratory of Digital Manufacturing,Wuhan University of Technology,Wuhan 430070,China;School of Mechanical and Electrical Engineering,Wuhan University of Technology,Wuhan 430070,China)

机构地区：[1]武汉理工大学湖北省数字制造重点实验室,武汉430070 [2]武汉理工大学机电工程学院,武汉430070

出　　处：《武汉理工大学学报》2022年第3期71-78,共8页Journal of Wuhan University of Technology

基　　金：国家自然科学基金青年基金(51505355);国家重点研发计划(2017YFC0703903-04)。

摘　　要：针对建筑工业化吊运装备运行安全预警问题,提出了一种基于多传感器的吊运装备运行过程障碍物检测方法。首先利用阈值筛选和层级聚类算法对毫米波雷达数据进行处理,滤掉无效目标;利用扩展卡尔曼滤波算法对目标进行跟踪;然后,利用改进的YOLOv4算法对采集到的障碍物数据集进行训练,从而实现障碍物的多目标检测。最后,采用决策级融合策略融合毫米波雷达与机器视觉信息。为验证所设计的融合方法,进行了不同施工环境下的测试实验。结果表明:该方法可实时检测跟踪前方障碍物,检测准确率达93.1%,比毫米波雷达与传统机器视觉融合的障碍物检测方法有更好的可靠性与鲁棒性。Aiming at the early warning problem of the operation safety of construction industrialized lifting equipment,this paper proposes a multi-sensor-based method for detecting obstacles during the operation of lifting equipment.First,use threshold filtering and hierarchical clustering algorithms to process millimeter wave radar data to filter out invalid targets;use the extended Kalman filter algorithm to track the targets;then use the improved YOLOv4algorithm to train the collected obstacle data sets,So as to achieve multi-target detection of obstacles.Finally,a decision-level fusion strategy is adopted to fuse millimeter wave radar and machine vision information.In order to verify the designed fusion method,test experiments under different construction environments were carried out.The results show that the method can detect and track obstacles ahead in real time,with a detection accuracy of 93.1%,which is more reliable and robust than the obstacle detection method fused with millimeter wave radar and traditional machine vision.

关 键 词：毫米波雷达 传感器融合 摄像头 YOLOv4 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]