Image Processing Based Vision Measurement Of Model Displacement In Shaking Table Test

Shaking table test is the most important and direct method to study the seismic performance of structures,and the model displacement response is an important physical quantity in the results of the shaking table test.The traditional displacement sensor applied in the shaking table test can basically meet the requirements of displacement measurement,but there are still some limitations and deficiencies.As a non-contact measurement method,the visual measurement method has many advantages compared with the traditional displacement sensor,and can effectively solve or make up for the limitation of the traditional displacement sensor.Therefore,this paper has compiled a vision-based measurement program for the displacement measurement in the one-way shaking table test.At the same time,a series of shaking table tests have been carried out for a comprehensive analysis of the accuracy of the existing displacement measurement methods.The main research content includes the following three parts:（1）In order to solve the limitation of traditional displacement sensors,such as LVDT,laser displacement meter and vibration pickup in the displacement response measurement of shaking table test,this paper compiled a multi-point displacement response measurement program of shaking table test structure based on python language and computer vision related algorithms.The circular and grid-shaped artificial targets are designed as image features to be pasted on the structure surface where the displacement measurement is required,and a monocular high-speed industrial camera is used to obtain the targets’ image sequence of the shaking table test process.Computer vision-related algorithms are applied to segment the image of the targets from complicated backgrounds,including binarization,morphological operations,edge detection,contour recognition,etc.By setting contour features as filters,the problems of multi-targets extraction under complex backgrounds is solved.For the extracted circular target image,the Otsu algorithm is applied to segment the key image region,so as to obtain a circular region with a stable boundary in the marker image area.Ellipse fitting and gray-scale centroid method are then used to locate the center of the artificial targets.The sub-pixel accuracy level corner recognition and positioning algorithm is applied to positioning the center of the square-shaped targets.After obtaining the coordinates of the center of the targets in the image coordinate,the world coordinates of the targets’ centers are obtained through coordinate conversion,then the structure displacement response is obtained.A series of shaking table tests have verified the reliability and accuracy of the visual measurement program.（b）Binocular 3D digital image correlation algorithm is used to track and locate the artificial targets,including binocular stereo vision,camera calibration,stereo matching and 3D reconstruction process.The three-dimensional coordinates of the measuring point in the world coordinate system are therefore obtained,thereby obtaining the structural displacement response.The above four vision-based measuring methods are simultaneously applied,including the three-dimensional digital image correlation method and the visual measurement program（three marking point positioning methods）,to measure the displacement response of the three-story steel frame electric shaking table test.An electric vibrating table with a highprecision grating ruler is used to apply seismic loads to the structure and at the same time serve as a reference standard for the measurement results of the structure’s displacement response.Sine waves and seismic waves were selected as the input signals of the shaking table,and the visual measurement results were compared with those obtained by acceleration sensors,displacement transducers,laser displacement meters and grating rulers.Compared with the displacement response reference standard,the minimum RMSE error value of the measurement result of the visual measurement program can reach 0.002.（c）In order to verify the effectiveness and accuracy of the visual displacement measurement program written in this paper in the large-scale structural shaking table test,a monocular industrial camera was used to collect images of artificial marker points and a large-scale cassette structure model shaking table test was carried out.The multi-point displacement measurement program is used for image processing to obtain the structural displacement response.The measurement results are compared with the results of the laser displacement meter and the vibration pickup,and the average error is within 5%.The validity and accuracy of the written measurement program and the adopted visual measurement method are therefore verified.