The Research On Optical Measurement System Of Stereo Vision Based On 3-D Digital Image Correlation

The optical measurement system of stereo vision based on 3-D digital image correlation with advantages of non-contract and high precision is a high-efficient measurement instrument,which is able to measure the shape and deformation of an object surface and is widely used in machining and manufacturing,virtual reality,reverse engineering.And it has become a research focus at home and abroad.Hence,it has very important theoretical value and practical significance to research on optical measurement system of stereo vision based on 3-D digital image correlation.The paper around the difficulties in the development of the image acquisition module,stereo calibration,epipolar rectification and stereo matching is discussed further.Firstly,according to the characteristic of the optical measurement system,develop a multi-function image acquisition module based on Labview which can make multiple cameras to capture simultaneously and adjust the corresponding camera parameters according to the experiments’ requirements.The basic principle of binocular stereo vision and the camera coordinate system as well as their homogeneous coordinate transformation are analyzed.And two camera model calibration methods are introduced detailedly.According to characteristics of the optical measurement system,the secondary calibration method using 2-D checkerboard plane template is proposed to improve the calibration accuracy.The imaging principle of parallel binocular stereo vision model and non-parallel binocular stereo vision model are analyzed particularly.Secondly,the key technologies in image matching are studied deeply from the following aspects:1 To overcome the disadvantages of low precision,large amount of calculation and low robustness of traditional stereo matching method such as local method and global method,a matching method based on Bayesian model and digital image correlation is proposed.This method uses the disparity to search space through building Bayesian model which can reach the purpose of no global optimization and provide a reliable disparity initial value for Newton-Raphson iteration.Compared with traditional stereo matching methods,the matching method based on Bayesian model and digital image correlation has considered the deformation of subset and adopted the subpixel localization algorithm to compute the disparity with subpixel precision which can improve the matching precision.2 The principle of image interpolation is analyzed detailedly,and to build the look-up table to interpolate the subpixel.3 A new 3-D coordinate calculation method has been proposed.This method searches the matching points on raw stereo images according to the transformational relation between raw and rectified image,which can overcome mistakes caused by rectification compared to using matching points on rectified images directly.4 An improved method to compute the initial displacement value has been analyzed that is named seed point method.Actually,this seed point method is an improved path dependence method which can reduce the amount of calculation greatly5 The least square method is adopted to fit the Cauchy or Green strain.Finally,several experiments have been operated to test the measurement results of the optical measurement system in the paper.