基于计算机视觉人行桥挠度影响线非接触式识别

A NON-CONTACT RECOGNITION FOR DEFLECTION INFLUENCE LINE OF FOOTBRIDGE BASED ON COMPUTER VISION

该文将便携式相机与无线传感器结合,开发了一种人致挠度影响线非接触式识别系统,避免了传统识别方法需要长时间阻断交通、耗费大量人力物力等不足,可用于运营状态下的桥梁影响线识别。通过便携式相机获取桥上行人行为,引入遮挡模型改进YOLO算法识别桥上行人,跟踪目标行人坐标变化得到行人位置信息,结合无线传感器得到的行人荷载作为结构输入数据。通过视觉识别技术跟踪结构行为获得人行荷载作用下的位移响应作为结构输出数据。根据结构输入输出数据反算人行荷载作用下桥梁挠度影响线。对初始影响线进行高阶滤波处理,消除环境和其他因素干扰,然后利用多项式分段拟合实测桥梁影响线进而得到具有准静态特性的挠度影响线,可为结构工程师准确高效地提供桥梁损伤检测依据。

This paper develops a non-contact recognition system of the influence line of human-induced deflection combined with the portable camera and wireless sensor, which avoids the shortcomings of traditional identification methods. It concludes of long-time blocking off traffic and of consuming a lot of human resources and material resources. The engineers can use this system to identify the influence line of bridges in operation. A portable camera obtains the pedestrian behaviour on the bridge. The occlusion model is introduced to improve the Yolo algorithm to identify the pedestrian on the bridge. Tracking the changes of the coordinates of the target pedestrian can obtain the pedestrian position information and the pedestrian load acquired by the wireless sensor is taken as the structural input data. Visual recognition technology is used to track the structural behaviour and the displacement response under pedestrian loading as the output data of the structure. According to the input and output data of the structure, the deflection influence line of bridge under pedestrian loading is calculated reversely.The initial influence line is processed by high-order filtering to eliminate the interference of environment and other factors. Then, the measured bridge influence line with characteristics of quasi-static characteristics is fitted using the polynomial segmentally, which can provide an accurate and efficient basis to detect the bridge damage for structural engineers.