Linear Structure Dynamic Small Offset Video Measurement Method With Monocular Camera

With the construction and operation of linear engineering structures such as highways and bridges,the traffic network in country has been gradually improved,which has effectively relieved traffic congestion and improved traffic mobility.Linear engineering structures have the characteristics of long length,wide distribution range,complex spatial environment,and long operating time,etc.Under the influence of natural and human factors,it is prone to produce cracks,diseases and other risk factors that damage the stability of the structure.Therefore,before the actual construction of linear engineering structures,shaking table dynamic load response tests are conducted on the structural models under certain scaling scales,to verify the overall stability of the structure and optimize the design of structural members.With the advantages of non-contact,high frame rate,high accuracy,low system complexity,large effective field of view,flexible deployment location,low space and economic cost,monocular camera high-speed video measurement technology is widely used in the study for dynamic load response monitoring of shaking table structural models,providing data basis for refined numerical simulation and theoretical analysis of structural models.At present,the main way to monitor the dynamic offset of the structure model is to deploy artificial signs at key nodes,but there are problems of low localization accuracy due to poor edge fitting of artificial signs,and it is difficult to robustly and efficiently monitor the region of interest where no artificial signs are deployed.To address the above-mentioned challenges,this thesis proposes a high-precision automatic identification and localization method for artificial targets and an EET-Hamming measurement method for natural targets of structures,which achieves robust and efficient measurement of structural dynamic small offsets in complex environments by monocular camera video measurement technology.The main work of this thesis is as follows:(1)Aiming at the problem of inaccurate positioning due to poor edge fitting of signs in manual sign measurement,this thesis constructs a high-precision automatic recognition and positioning method based on manual targets.First,a highly robust black-and-white circular sign is designed by combining the characteristics of high-speed video images with low luminance,and an automatic target recognition method based on the morphological processing of Region of Interest(ROI)window is established,in order to reduce the human intervention in the recognition process and enhance the automation of image processing;second,the overall least squares fitting method is used to edge fitting and calculate the artificial signs circle coordinates based on the optimized Levenberg-Marquardt(LM)method;finally,the matching tracking strategy based on the target neighborhood blocks is used to obtain the dynamic offset results of the time-series images.The experimental results show that the accuracy of the proposed circle center identification fitting localization method in this thesis is 9.1% better than that of the traditional localization algorithms Canny and Sobel.(2)Aiming at the problem that linear structural natural targets are difficult to track robustly and efficiently,this thesis constructs an EET-Hamming-based monocular video measurement method.First,the Edge-enhanced Transform(EET)method is used to enhance the image and the Otsu binarization method is used to remove the redundant features in the image;second,the minimum Hamming distance matching method is proposed to efficiently match the images before and after deformation;finally,a nine-point least-squares polynomial surface fitting method is proposed to obtain high-precision subpixel coordinates.The experimental results show that the RMSE of the proposed EET-Hamming method for monocular video measurement of high and low contrast targets is 0.13 mm and 0.07 mm,respectively,and the matching efficiency is improved by 13.9% compared with the Normalized Cross-Correlation(NCC)method.(3)In the dynamic offset experiment of ramp bridge structure,to address the problem that it is difficult to obtain the displacement change of the plate rubber bearing structure by traditional sensors,this thesis extracts the dynamic offset information of the structure from the video by deploying artificial signs and using automatic identification and positioning algorithm.The experimental results show that the maximum RMSE between the measurement results of this thesis and those of commercial instruments is 0.21 mm.In the dynamic offset experiments of cable-stayed bridge structure,to address the problem that traditional sensors are easy to cause mass effect when deployed in lightweight cable-stayed bridge structure,this thesis uses the proposed EET-Hamming method to measure the slight vibration of shaking table and the dynamic offset information of each monitoring point based on the structural natural target,the location of the easy damage points of the structure is determined.Finally,the obtained offset information data and the corresponding accelerometer data are analyzed by spectrum analysis,and the experimental results further verify the effectiveness of the proposed method.