The Study Of Binocular Stereo Vision 3D Reconstruction Based On Digital Speckle

The passive stereo vision is one of the common means to obtain three-dimensional information of the object,which has great significance and broad application prospect in the areas,such as machine vision,industrial inspection,medical diagnosis.Because the common grating projection 3D measurement method requires complex phase unwrapping process,its scope of application is limited,3D digital speckle correlation method has been rapid development,it is based on binocular stereo vision with the aid of digital speckle,additional features of the object surface is added,the precise dense disparity map is obtained after stereo matching,finally the three-dimensional coordinates of the object is got based on camera calibration results.Firstly,in this paper,the measurement principle of parallel binocular stereo vision is studied,the relationship between the parallax and the three-dimensional information of the object is deduced.The camera distortion model and the camera calibration method are studied,the conversion relationship between the world coordinate system and the image coordinate system is determined,the image coordinates of each corner of the board are obtained by means of the vision and motion kit of Lab VIEW,based on the known world coordinates,the external and internal parameters of two cameras are solved by using the two-step calibration method of Tsai,compared with value of the standard parameter,the measurement result is relatively satisfied.Secondly,the digital speckle algorithm of the computer generated simulation is studied in detail,the algorithm of local position random digital speckle is deduced on the basis of the algorithm of global position random digital speckle,global position random digital speckle and local position random digital speckle are generated by using MATLAB,at the same time the speckle properties and image size for the influence of the speckle image quality of object is discussed.Finally,since the correlation of the left and right images is the most important part of the stereo matching,in this thesis,the normalized co-variance correlation algorithm is studied,then the parabolic sub-pixel interpolation algorithm is introduced,the accuracy of stereo matching is improved.The experiment hardware platform is built,the speckle image of the object is collected,the matching results of face mold and conical object are given,the relatively satisfied bushy gray disparity map is obtained,then the three-dimensional coordinates of the object is measured based on the camera calibration,the high value of feature points and the measurement results of vernier caliper are basically identical.