Stereovision Based Digital Image Correlation Method For Monitoring Three Dimensional Deformation

In recent years, there has been increasing attention to computer vision technologyin all walks of life, such as aerospace, engineering,mechanical measurements,automobile license plate recognition,3D reconstruction of medical science, etc. we tryto research it to apply to the structural deformation monitoring. With the purpose ofmeasuring3D deformation of structure in civil engineering, the paper do someselections and improvements on binocular stereo vision technology-related matchingand searching algorithms, the displacement and strain of any small area on the structuresurface can be measured from programming.The paper chooses high precision-calibration algorithm: three-step method, usesfeature matching and digital image correlation method(DIC) respectively on stereomatching and sequential matching, and the results are reasonable. In the sequentialmatching, proposing ROI-based search method to improve search stategy. The searcharea can be reduced, which reduces100times of cumulative calculation. The researchesof the paper are as follows:Firstly, the relevvant principles of binocular stereo vision and model solutionmethod are analyzed and studied respectively. The images are rectified by epipolarrectification. Using the histogram specification enhance image,using double wienerfiltering and mean filtering denoising. In this paper, opening operation is adopted toimprove the light equalization and solve the miss matching problem. The paper selectsfeature guided Bayesian estimation algorithm as our stereo matching method which cangenerate dense disparity map quickly. Choosing the ROI-based search strategy toimprove the method of DIC on sequential matching. Three pixel interpolation methodsare analyzed, and the bilinear interpolation algorithm is adopted for pixel interpolation.Matching template is proposed and adjusted to the actual regional characteristics, thusthe most reasonable(obvious characteristics and small amount of calculation) areaaround the point of interest can be manually selected.Using Newton-Raphson methodwhose iterative initial value is integer pixel search results above to work out thedisplacement and strain exactly.Secondly, the paper discuss the calibration algorithms, emphatically analyzeZhengyou Zhang classic method and the three-step method which is proposed recently, we choosethe three-step method by comparing both reprojection error and given length reconstruction error ofthe two methods. A deformation measurement system is developed according to themethod of this paper using MATLAB program, and using the translation machine testthe displacement measurement accuracy and feasibility of the system. Using tensile experiment of speckled plate test the accuracy of strain measurement. In the end, thedisplacement of a three-story framework model is measured, and the availability of themethod in civil engineer is validated.