3D Reconstruction Method Of Binocular Stereo Vision Based On General Digital Cameral Image

Binocular stereo vision theory which proposes a solution for scene three-dimensionalinformation restoration by using two-dimensional images is a hot research field of3Dreconstruction. Based on binocular stereo vision theory, this paper demonstrates an automaticmethod to extract3D coordinates of the key points in images taken by ordinary digital camera.Throughout the course of the study, the paper discussed in detail in parts of stereo matching ofdigital images, digital camera calibration,3D reconstruction etc., and mainly completed thefollowing works:(1) Image pre-processing: Using adaptive median filter and Wallis sharpening respectivelyon stereo pair pretreatment, the purpose of which is to eliminate noise interference, enhanceimage edges and detail information.(2) Stereo matching: On the basis of SIFT algorithm, combined with homography and polaras double constraint conditions for constraint matching, and applied RANSAC algorithm toeliminate the error matching, the paper realizes a coarse to fine, sparse to dense matchingsolution to improve the matching speed and accuracy.(3) Camera Calibration: The paper uses Zhang Zheng You plane calibration method,which mainly uses the world coordinate of the given physical point and the pixel pointcoordinate of the template image (combining Harris corner detection and quadric surface fittingto extract the subpixel level coordinates of checkerboard Angle points) to calculate the camerainner parameters according to camera imaging model.(4)3D reconstruction: Using the pixel coordinates of feature points obtained by stereomatching and inner parameters acquired by camera calibration, the paper calculates thethree-dimensional coordinates of stereopair feature points according to the principle of SFM, andputs forward a coordinate comparison method to validate the accuracy of3D coordinates andfurther determine the feasibility of the scheme.Each chapter experimental results show that every part can reach the expected effect and thewhole system can get3D information of space points quickly and precisely. These results alsostrongly prove the feasibility of the proposed method. For ordinary users whose cameraownership rate is increasing fast, this paper provides a new solution to restore3D information of space scene based on digital images. For the professional fields requiring3D model such asdigital city, this paper also offers a good alternative reference solution!...