双目视觉感知下的公路限高实时预警算法  

Real-time monitoring algorithm of highway height limit via binocular vision

作　　者：李宝军[1] 薛炯 刘泽阳 王小超 肖志鹏 LI Baojun;XUE Jiong;LIU Zeyang;WANG Xiaochao;XIAO Zhipeng(School of Automotive Engineering,Dalian University of Technology,Dalian 116024,China;School of Mathematical Sciences,Tiangong University,Tianjin 300387,China;Beijing Smarter Eye Technology Co.,Ltd.,Beijing 100023,China)

机构地区：[1]大连理工大学汽车工程学院,辽宁大连116024 [2]天津工业大学数学科学学院,天津300387 [3]北京中科慧眼科技有限公司,北京100023

出　　处：《重庆理工大学学报(自然科学)》2023年第11期73-81,共9页Journal of Chongqing University of Technology:Natural Science

基　　金：国家自然科学基金项目(52172382);国家重点研发计划项目(2021YFB3300602)。

摘　　要：针对车辆运动工况下对种类繁杂的公路限高设施测高难的问题,提出一种双目相机限高目标实时监测算法。通过分析限高场景及设施,采集并构建大规模多场景的公路限高数据集;采用基于深度学习的检测和跟踪算法获取目标区域;利用多步多尺度的滤波算法对双目视差数据进行处理,得到高置信度的限高设施视差数据,进而计算距离及高度信息;最后采用鲁棒性输出策略对计算得到的测距和测高数据进行输出预警。通过多场景限高测试,目标检测精度为98.77%、漏检率为0.87%、误检率为0.36%,在60 m内的平均测高误差小于4%。实验结果表明该算法可以实现公路限高目标的自动检测,能够实时鲁棒地输出测距、测高和预警信息。To confront the difficulties for moving vehicles in measuring the actual heights at various highway height limit facilities,this paper proposes a binocular camera height limit target real-time monitoring algorithm.First,an analysis of the height limit scenarios and facilities is made and a large-scale multi-scenario highway height limit dataset is built.Second,the target region is obtained by the detection and tracking algorithm based on deep learning.The binocular disparity is processed by a multi-step and multi-scale filtering algorithm to obtain the high-confidence height limit facility disparity data and thus calculate the distance and height information.Finally,the robust output strategy is adopted and alarms will be activated based on the real-time collected distance and height results.After multi-scenario height limit experiments,the target detection accuracy stands at 98.77%,the missed detection rateat 0.87%,the false detection rate at 0.36%,and the average height measurement error within 60 m below 4%.The results show that our algorithm achieves automatic detection and computation of highway height limit targets,as well as real-time and robust delivery of distance and height measurements and early warnings.

关 键 词：公路限高 多步多尺度滤波 鲁棒性输出 自动检测 

分 类 号：U461[机械工程—车辆工程]