Study On Mechanical Properties Of Thin-Walled Steel Members Based On Binocular Vision

Structural deformation measurement is a direct means to monitor structural safety and evaluate structural performance.With the rapid development of science and technology,traditional measurement methods have been replaced by computer vision technology.Nowadays,computer vision technology is changing with each passing day,among which binocular stereo vision technology has become a hot spot in computer vision research,and is widely used in many fields such as 3D surveying,robot autonomous navigation,scene reconstruction and object recognition.In the actual engineering measurement,the three-dimensional digital image correlation method is usually used to measure the structural deformation.Compared with the traditional contact deformation measurement technology,binocular stereo vision technology has the advantages of non-contact,low cost and full-field measurement.Compared with the optical non-contact deformation measurement technology based on laser and optical interference,binocular stereo vision technology has the characteristics of lower environmental requirements and simple experimental measurement equipment,and is widely used in the field of engineering experiment research.Through theoretical research,a binocular test system is set up,and the speckle quality of the test is evaluated based on MATLAB platform.On this basis,the axial buckling test of Csection cold-formed thin-walled members is carried out,aiming at studying the advantages and accuracy of 3D digital image correlation method in the buckling deformation test of thinwalled steel.The main research contents are as follows:(1)In the part of theoretical basis,firstly,the calibration principle and two calibration methods of camera stereo calibration are introduced,and the mapping relationship between calibration coordinate systems is deduced.Secondly,stereo correction is introduced,mainly epipolar correction,which is the key to stereo matching of subsequent images.Thirdly,the methods of image preprocessing,such as grayscale processing,geometric transformation,image enhancement and image denoising,are introduced.Then,regarding stereo matching,it introduces the different classifications of stereo matching and the four basic steps of stereo matching,namely matching cost calculation,matching cost aggregation,disparity calculation and disparity optimization.About 3D reconstruction,general point cloud computing and parallel vision point cloud computing are introduced respectively.(2)Based on MATLAB platform,the speckle quality of three kinds of speckle images is evaluated by the average gray gradient and the second derivative of the average gray gradient,and the high-quality speckle is screened out,which can improve the search and matching accuracy.Moreover,the high-quality speckle is applied to the observation surface of the specimen,which improves the efficiency and accuracy of the test.(3)The mechanical properties of 6 standard specimens cut from flange and web before and after cold working and the bearing capacity of 16 cold-formed thin-walled steel specimens with C-section rolled under axial compression were tested.The deformation,failure mode and stability of members under axial compression are studied,and the traditional measurement method,self-made 3D measurement software and digital image correlation method based on binocular vision are compared.The measurement error of traditional measurement method and self-made 3D measurement software is 0.04～0.2mm,and the measurement error of traditional measurement method and digital image correlation method based on binocular vision is 0.03 mm,which proves the feasibility of binocular vision method.