Computer Vision-Based Dynamic Displacement Monitoring Of Shaking Table Tests Of Structures

In recent years,with the continuous development of 5G communication technology and image acquisition equipment,computer vision-based structural measurement technology has been widely used in the field of Structural Health Monitoring(SHM)in civil engineering,in which the dynamic displacement response of the structure,as one of the most important information,has far-reaching significance for its research.Since the traditional contact displacement sensor measurement method has the disadvantages of tedious multi-point measurement,complicated transmission line and inconvenient fixed installation,the vision-based non-contact displacement measurement method can make up for the shortage of traditional contact displacement sensors.Although the monocular vision method can monitor the in-plane displacement of a structure in two dimensions with high accuracy,it still cannot monitor the multi-directional displacement of a structure at the same time.Therefore,the introduction of binocular stereo vision technology based on the principle of parallax can complete the dynamic monitoring of the three-dimensional displacement of the structure.To this end,this paper uses Python language to write a monocular-based structural displacement visual measurement program and a binocular-based structural 3D displacement visual measurement program based on existing research.And the stability and accuracy of the algorithm are verified by structural shaking table test.The main research includes the following parts:(1)Security cameras and digital cameras are applied in visual displacement monitoring of structural shaking table test.Cam Shift tracking algorithm based on color histogram and LK pyramid optical flow tracking algorithm based on Shi-Tomasi corner detection are respectively adopted to realize dynamic displacement monitoring of five-storey steel frame structure model.The results obtained by the two algorithms are compared with those obtained by the pull-line displacement meter.The results show that the optical flow method based on corner detection is superior to the Cam Shift tracking algorithm based on color histogram in both accuracy and application conditions.(2)LK pyramid optical flow tracking algorithm based on Shi-Tomasi corner detection.By using a consumer digital camera to collect the vibration video of a three-layer plane irregular RC frame isolation structure with a scale of 1:4 under the action of ground motion,Shi-Tomasi corner tracking in the ROI region at the beam-column node is carried out.Then the real displacement response of each layer of the structure under earthquake is obtained.By comparing the visual measurement results with the data of the drawstring displacement meter,the displacement time-history curves obtained by the two measurement methods are i n good agreement.The maximum error of peak point is 2.17%.The correlation coefficient ρ and the normalized mean square error NMSE are both above 0.9.The information in the frequency domain of the time domain signal is obtained by Fourier transform.The visual measurement results in the frequency domain are basically consistent with the measurement of the pull-line displacement meter,and the visual measurement method can identify the structural vibration frequency more accurately.It shows that this visu al measurement method can be used for multi-point dynamic displacement monitoring of large structure shaking table test.(3)In view of the high precision measurement of the out-of-plane displacement of the structure cannot be completed simultaneously with monocular vision,the dynamic measurement program of the three-dimensional displacement of the structure is written based on the principle of binocular stereo vision.In the process of stereo matching,the number of feature points extracted by SIFT stereo matching algorithm is large,resulting in large amount of calculation,long operation time,unable to measure in real time,and easy to appear mismatching deficiencies.In this paper,based on the existing research,the SIFT stereo matching algorithm is improved,that is,the Shi-Tomasi operator and SIFT algorithm are combined to extract Shi-Tomasi corner points on the image and carry out stereo matching between the left and right images,so as to achieve the three-dimensional coordinate reconstruction of Shi-Tomasi corner points.(4)The three-dimensional displacement measurement program based on binocular stereo vision is verified by shaking table test of five-story steel frame structure model.Firstly,based on Zhang Zhengyou calibrati on method,the binocular camera calibration is realized by using Matlab toolbox.Secondly,Bouguet algorithm is used to carry out stereoscopic correction of the structural vibration image sequence collected by binocular camera.Finally,the binocular visio n three-dimensional displacement measurement method is used to complete the in-plane displacement monitoring of the model structure shaking table test.The accuracy of binocular stereo vision based out of plane displacement measurement method is analyzed a nd evaluated in time and frequency domain.The results show that the displacement time-history curves of the two measurement results are in good agreement under each working condition,and there is a slight difference at the peak displacement.Fourier tran sform is applied to the displacement time history data,and the peak frequency of the structure measured by the two methods is basically the same,but the amplitude is slightly different.In order to further verify the robustness and applicability of this visual measurement method,dynamic displacement monitoring of model structure shaking table test with binocular camera under different baseline distance and deflection Angle under earthquake was completed.The results show that the correlation coefficient ρ is above 0.9 and NMSE is above 0.8 compared with the results of the wire-line displacement meter.The visual measurement method can complete the dynamic displacement monitoring of the structure under earthquake with high precision under the condition tha t the camera baseline distance is variable and the camera deflection angle is different.