Abstract: Bridge displacement influence line plays an important role in bridge condition assessment. Traditional bridge displacement influence line identification methods mostly rely on contact sensors, which have problems such as low efficiency and obstruction of traffic. In order to realize convenient and intelligent recognition of bridge displacement influence line, a bridge displacement influence line recognition method based on machine vision and nonlinear frequency modulation signal decomposition was proposed. In the proposed method, the phase-based video motion magnification (PVMM) algorithm was first used to select the phase change of the frequency sub-band of interest in the video and amplify it. The one-dimensional row-gradient domain-guided image filter (Row GDGIF) was used to finely remove the bilateral, artifacts and other noises in the amplified sub-band, and then the motion-amplified video was synthesized. Then, the sub-pixel edge detection technology was used to identify the edge of the bridge, and vibration displacement response of the bridge under moving vehicle load was obtained. Finally, considering the time-varying characteristics of the bridge vibration displacement response frequency, the nonlinear frequency modulation signal decomposition technique was introduced to separate the static and dynamic components of the bridge vibration displacement, realizing bridge displacement influence line identification at last. The applicability and robustness of bridge displacement influence line identification method based on nonlinear frequency modulation signal decomposition was verified by the combination of finite element numerical simulation. The experimental research on the identification of bridge displacement influence line was further conducted. Based on machine vision, the vibration displacement of the bridge under different vehicle speeds was extracted, and the nonlinear frequency modulation signal decomposition method was used to identify the bridge displacement influence line. The results show that the proposed method can effectively identify the bridge displacement influence line and have relatively high accuracy.

Key words: bridge engineering; bridge displacement influence line; machine vision; one-dimensional row gradient domain guided filtering; nonlinear frequency modulation signal decomposition

摘要： 桥梁位移影响线在桥梁状态评估中具有重要的作用。传统桥梁位移影响线识别方法大都需要安装接触式传感器，存在效率低、阻碍交通等问题。为实现便捷、智能的桥梁位移影响线识别，提出了基于机器视觉和非线性调频信号分解的桥梁位移影响线识别方法。该方法首先利用相位运动放大（PVMM）算法，选取视频感兴趣频率子带的相位变化并进行放大，采用一维行梯度域引导滤波器（Row GDGIF）对放大后该子带中的双边、伪影和其他噪声进行精细去除，合成运动放大后的视频；然后，利用亚像素边缘检测技术识别桥梁边缘，得到移动车辆荷载作用下的桥梁振动位移响应；最后，考虑桥梁振动位移响应频率的时变特性，引入非线性调频信号分解技术，分离出桥梁振动位移中的静态分量和动态分量，最终实现桥梁位移影响线识别。结合有限元数值模拟，验证了基于非线性调频信号分解的桥梁位移影响线识别方法的适用性和鲁棒性。进一步开展了梁桥位移影响线识别试验研究，基于机器视觉提取了不同车速情况下的桥梁振动位移，并利用非线性调频信号分解方法对桥梁位移影响线进行了识别，结果表明，方法可以有效识别桥梁位移影响线并具有较高精度。

关键词: 桥梁工程；桥梁位移影响线；机器视觉；一维行梯度域引导滤波；非线性调频信号分解