| Binocular vision is widely used in many fields of industry,because binocular vision can realize the conversion between two-dimensional information and three-dimensional information,so binocular vision is called the popular direction in the field of computer vision.In this paper,through single-target calibration,double-target calibration,stereo correction and stereo matching of the camera,the relevant parameters of the camera are obtained,and the weld image coordinates are obtained by correlation processing.The main content of the thesis includes:(1)The theory related to camera is introduced.The homogeneous coordinate system,the imaging principle of the Ideal Camera,the conversion from the world coordinate system to the image coordinate system and the imaging principle of the real camera are introduced.(2)Stereo Vision calibration is mainly to calibrate the camera to find out the camera internal and external parameters,including single camera calibration,binocular camera calibration.The camera calibration adopts checkerboard calibration method of Zhang Zhengyou.The inner parameters of the camera are obtained by the monocular camera calibration,and the outer parameters between the two cameras are obtained by the binocular camera calibration.At the same time,the reprojection Matrix is obtained by stereo correction.(3)Stereo matching.the similarities and differences of feature extraction algorithm SIFT and SURF are analyzed,and the disparity principle is expounded.(4)preprocessing of weld image.Mainly through the image morphology processing,such as erosion and expansion,the processed image uses the median filter to eliminate the noise,carries on the binarization processing to the filtered image,then uses the thinning algorithm to carry on the centerline extraction to the Binarization Weld,to achieve the extraction of weld feature points.(5)according to the Feature Points of weld image processing combined with the reprojection Matrix and the principle of binocular visual Parallax,the two-dimensional coordinates of the image are restored to three-dimensional coordinates. |