基于多目标跟踪的桥梁车辆荷载动态监测方法  

Dynamic Monitoring Method of Large Bridge Vehicle Loads based on Multi-target Tracking

作　　者：张建功 龚凯 郭翠翠 ZHANG Jian-gong;GONG Kai;GUO Cui-cui(Hubei Communications Yunchu Construction Management Co.Ltd.,Shiyan 442000,China;Hubei Communications Investment Intelligent Detection Co.Ltd.,Wuhan 430000,China;China Railway Bridge Science Research Institute Co.Ltd.,Wuhan 430000,China)

机构地区：[1]湖北交投郧楚建设管理有限公司,十堰市442000 [2]湖北交投智能检测股份有限公司,武汉市430000 [3]中铁大桥科学研究院有限公司,武汉市430000

出　　处：《公路》2024年第6期107-116,共10页Highway

摘　　要：桥梁车辆荷载的动态分布对桥梁健康状态的监测和分析具有重要意义,传统的车辆荷载监测方法难以适用于跨度大、结构复杂的桥梁,并且不能实时监测。因此,针对传统车辆荷载监测方法的局限性,在利用动态称重系统的基础上,提出一种基于机器视觉的车辆荷载动态监测方法,能够应用于大跨度桥梁。首先在桥头布置动态称重系统对经过的车辆采集重量数据,同时利用计算机视觉系统同步采集称重车辆的图像并分配ID;车辆在沿桥行驶的过程中,为了计算车辆在桥面的位置,首先进行图像透视畸变校正,建立图像坐标与真实桥面坐标的映射关系,然后通过布置在桥梁上的视觉检测系统采用SSD算法识别沿桥相机视野内的车辆,并记录车辆在桥面上的位置信息;再根据多目标跟踪算法实现车辆目标在相邻时刻的ID关联;如此迭代,即可获取车辆在全桥的移动轨迹,并可根据车辆ID追溯其重量信息,以此达到实时监测桥梁车辆荷载的目的。为了验证所提出方法的可靠性,在武汉鹦鹉洲长江大桥进行了实验验证,结果显示,在不同天气下车辆识别的精确度为96.3%,多目标跟踪结果的ID跳变率为3.4%。实验结果表明监测效果稳定、可靠,能够满足大跨桥梁全桥面车辆荷载动态分布情况监测与记录的要求。Dynamic distribution of vehicle loads on bridges is important for monitoring and analysis of bridge health status.Traditional vehicle load monitoring methods are difficult to be applied to bridges with large spans and complex structures,as well as incapable to monitor in real time.It is therefore proposed in this paper,based on the use of weight in motion system,a machine vision-based dynamic monitoring method for vehicle loads,which can be applied to large span bridges,to address the limitations of traditional vehicle load monitoring methods.First,a weight in motion system is arranged at the bridge to collect weight data from passing vehicles,while a computer vision system is used to synchronize the images of the weighed vehicles and assign IDs.In order to calculate the position of the vehicle on the bridge during the vehicle's travel along the bridge,the image perspective distortion correction is first carried out to establish the mapping relationship between the image coordinates and the real bridge coordinates,and then the SSD algorithm is used to identify the vehicle within the field of view of the camera along the bridge through the vision detection system arranged on the bridge to save the position information of the vehicle on the bridge.Then the ID correlation of vehicle targets at adjacent moments is realized according to multitarget tracking algorithm.Iteration is repeated,and the moving track of the vehicle in the whole bridge can be obtained,and its weight information can be traced according to the vehicle ID,so as to achieve the purpose of real-time monitoring of the bridge vehicle load.Experiment validation is conducted at Wuhan Yingwuzhou Yangtze River Bridge in order to verify the reliability of the proposed method.The results show that the accuracy of vehicle identification under different weather is 96.3%and the ID variation rate of multi-target tracking results is 3.4%.The experimental results show that the monitoring effect is stable and reliable,and can be satisfied with the requirements of

关 键 词：桥梁监测 车辆荷载识别 计算机视觉 深度学习 多目标跟踪 

分 类 号：U441.2[建筑科学—桥梁与隧道工程]