基于计算机视觉的结构多测点位移与振动特性非接触测量方法  

Noncontact Measurement of Multipoint Displacement and Vibration Properties of Structure with Computer Vision Technology

作　　者：董晨阳 田永丁 向晖 李东华 周成龙 DONG Cheng-yang;TIAN Yong-ding;XIANG Hui;LI Dong-hua;ZHOU Cheng-long(China State Railway Investment Construction Group Co.,Ltd.,Beijing 102601,China;School of Civil Engineering,Southwest Jiaotong University,Chengdu 610031,China)

机构地区：[1]中建铁路投资建设集团有限公司,北京102601 [2]西南交通大学土木工程学院,成都610031

出　　处：《科学技术与工程》2024年第21期9104-9110,共7页Science Technology and Engineering

基　　金：四川省科技计划(2023NSFSC0893);中央高校基本科研业务费(2682022CX077)。

摘　　要：位移是衡量结构舒适性与安全性的一个关键参数,但传统接触式方法测量效率低、成本高且难以精确测量稠密测点位移。为解决以上问题,提出了基于计算机视觉与自适应变分模态分解的结构多测点位移和振动特性非接触测量方法。该方法利用高速摄像机获取冲击力锤作用下悬臂梁振动图像,通过有标靶和无标靶特征检测算法,获取结构在冲击载荷下的动态位移曲线,进一步提出了基于变分模态分解和希尔伯特变换的结构振动频率、阻尼比高精度识别方法。所提出方法利用悬臂梁非接触振动测量试验进行了验证,结果表明:通过圆心检测算法和图像特征检测算法提取位移的最大测量误差分别为0.093 mm和0.046 mm,证明了图像特征检测算法的位移提取精度更高;同时,利用变分模态分解方法对测量的位移信号进行了分离并识别结构振动特征,采用圆心检测和图像特征提取算法获取位移识别的固有频率、阻尼比基本一致。研究成果可为突发事件下桥梁结构多点位移高精度监测与振动特性识别提供一种新的方法,具有成本低、使用方便和测点布设灵活等优势。Displacement is a key parameter to evaluate the serviceability and safety of structures,but the traditional contact measurement method has the disadvantage of low efficiency and high cost,and it is difficult to accurately measure the dense displacements.To solve the above problems,a noncontact measurement method for the dynamic displacement and vibration properties of structures with computer vision technology and adaptive variational modal decomposition was proposed.A high-speed camera was employed to capture the motion images of a cantilever beam structure under the excitation of an impact hammer,and target detection and feature extraction algorithms without targets were developed to obtain the dynamic displacements of the structure under the impact load.Furthermore,a high-precision identification method for the vibration frequency and damping ratio of a structure based on the variational modal decomposition and the Hilbert transform was proposed.The effectiveness of the proposed method has been validated by the noncontact vibration measurement tests of a cantilever beam.The results show that the maximum measurement errors of displacement extracted by the circular center detection algorithm and the image feature detection algorithm are 0.093 mm and 0.046 mm,respectively,which proves that the displacement extraction accuracy of the feature detection algorithm is more accurate.Meanwhile,the variational modal decomposition method is used for the separation of the measured displacement signals and the identification of the structural vibration properties.It is seen that the identified natural frequencies and damping ratios by the circular target detection are almost the same as the results obtained from feature extraction algorithms.The research results can provide a new method for high-precision multipoint displacement monitoring and dynamic identification of bridge structures under unexpected events,and it has the advantages of low cost,ease of use,and flexible choice of measurement points.

关 键 词：位移测量 动力识别 非接触测量 计算机视觉 变分模态分解 

分 类 号：TU973.2[建筑科学—结构工程]