交通视频辅助的桥梁动态称重方法研究

为了进一步提升既有的桥梁动态称重技术，提出一种交通视频辅助的新型桥梁动态称重方法。首先介绍基于深度神经网络的计算机视觉目标检测技术和一种计算机视觉坐标转换方法，实现从交通监控视频中实时地探测与定位桥上行驶的车辆和车轴。然后引入桥梁应变分解方法和应变影响面识别方法，建立车重、车辆位置与桥梁应变之间的映射关系，从而建立一种综合利用时间和空间冗余信息对车辆进行称重的方法。该方法构建超定的影响面加载方程组，使用最小二乘法求解该方程组以得到桥上行驶车辆的轴重和总重。最后总结出一套交通视频辅助的桥梁动态称重方法框架。为验证以上方法，在某连续大箱梁桥的缩尺模型以及实桥上进行试验。试验包含单车、双车、跟车、并行、直行、变道、匀速、变速等复杂交通工况。模型试验结果表明：该方法的车辆总重识别误差均值为-2.02%，标准差为4.77%；车辆轴重的识别误差均值为4.77%，标准差为17.50%。实桥试验结果表明：该方法的车辆总重识别误差均值为0.21%，标准差为1.53%；车辆轴重的识别误差均值为-3.59%，标准差为42.67%。除此以外，所提出的方法还可用于识别桥上车辆的数量、类型、轴数、实时位置、运动轨迹、行驶速度等多粒度交通信息。

Research on Traffic-video-aided Bridge Weigh-in-motion Approach

In order to improve the existing bridge weigh-in-motion (BWIM) technique, a novel traffic-video-aided BWIM methodology is proposed. First, an object detection method based on the deep neural network and a coordinate transformation method is introduced. They were used to detect and locate vehicles and axles on the bridge in real time. Then, a bridge strain decomposition method and a method to identify the strain influence surface of bridge structures were proposed to establish the mapping relationship between axle weight, axle position, and vehicle-induced static bridge strain. Subsequently, a method that comprehensively uses temporal and spatial redundancy information to identify the axle weight and gross weight of vehicles was proposed. The method first constructed an overdetermined equation set of influence surface and used the least square method to solve the equation set then to obtain the axle weight and gross weight of vehicles on the bridge. The framework of the novel BWIM approach was summarized and verified through scale model experiments and field tests on a continuous large box girder bridge. The experiments included various traffic scenarios such as single vehicle, double vehicle, following vehicle, parallel vehicle, straight path, curved path, constant speed, and variant speed. Results of the model experiments show that the mean and the standard deviation of the error for identifying gross vehicle weight (GVW)is -2.02% and 4.77%.The mean and the standard deviation of the error for identifying axle weight (AW) is 4.77% and 17.50%.Results of the field tests show that the mean and the standard deviation of the error for identifying GVW is 0.21% and 153%. The mean and the standard deviation of the error for identifying AW are -3.59% and 42.67%. In addition, the method can also be used to identify the traffic information such as the number, type, number of axles, real-time position, trajectory, and speed of vehicles on the bridge.