Target-Free Structural Vibration Signal Extraction Method Based On Computer Vision

Structural vibration monitoring is an important component in the field of structural health monitoring,and it is the main means for structural dynamic performance evaluation and health analysis.The traditional contact measurement method requires sensors to be arranged on site,which has the disadvantages of complicated operation,single measurement point,additional load generation,and difficulty in installation of high-rise structures.In recent years,structural vibration monitoring methods based on video images have become a hot spot in structural health monitoring research.Compared with traditional sensor measurement methods,they have the advantages of non-contact,nondestructive,easy to operate,and can realize long-distance and large-scale multi-point monitoring.It is generally necessary to set targets on the structure,which is not a real "non-contact measurement".In addition,this kind of method has many defects,such as high requirement of equipment,easy to be affected by environment background and light,which limits its application in practical engineering.This paper has carried out related research on these problems existing in the existing methods,and provides a new and effective vibration monitoring method for the structure.This article first introduces the OpenCV development tools and completes the development environment.Aiming at the problems of expensive and inconvenient carrying of cameras and industrial cameras,a smartphone is used as an observation device and using FFmpeg programming to obtain a video time stamp solves the problem of difficulty in solving the time domain caused by the variable frame rate of mobile video.In-depth study of Zhang Zhengyou's camera calibration method,detailed description of the calibration process and key points,a new method of placing the calibration plate is proposed,which effectively solves the problem that the calibration plate is too small in the camera's field of view when the structure is far away.The LK optical flow method is used to realize the motion tracking of points,and the image pyramid is introduced to improve it,the improved pyramid LK optical flow algorithm is used.Considering the tracking effect and the number of tracking points,three types of feature points were selected as tracking points,and feature point detection tests were conducted on a variety of vibration structures of various types and backgrounds,which showed that the number and location of available points in most scenarios could meet the measurement needs.Based on the full text study,a new method of structural vibration signal extraction based on computer vision developed in this paper,without the need to arrange targets on the structure surface,using a smartphone as an observation device,the camera calibration process takes into account both image distortion and camera pose,and uses an improved pyramid LK Optical flow algorithm realizes point motion tracking,and uses OpenCV to compile a complete set of programs from video acquisition and decomposition,calibration parameter solution,distortion correction to feature point detection,point motion tracking,and displacement result output.Finally,the method is verified by the vibration test of the three-layer frame structure in the laboratory.The test results show that the method can achieve a high measurement accuracy.The error of the first two frames of the measured structure natural frequency is less than 1% compared with the results of the acceleration sensor.At the same time,it has the advantages of low equipment and environment requirements,and does not need complex image processing.The method has a good application prospect.