Research On Interactive Three-dimensional Reconstruction Technology Based On Binocular Vision

In modem society,information technology is developing rapidly,people are increasingly demanding visual processing technology,and interactive imaging technology is becoming more and more widely used.It plays a role in the restoration of cultural relics and historic sites,binocular ranging,medical and health,and unmanned driving An important role,in which interactive 3D reconstruction technology is the core technology in interactive imaging.In binocular interactive 3D reconstruction technology,feature point detection and matching is the key link,but the interactive 3D reconstruction based on binocular vision has a high mismatch rate in feature point detection and matching,resulting in low accuracy of 3D reconstruction results.There is a problem of missing pixels on the surface of the reconstructed object.These shortcomings reduce the visual experience of 3D reconstruction.In order to solve the problem of low matching efficiency and partial distortion of the reconstruction results in most current binocular interactive 3D reconstructions,this thesis combines SOBEL edge detection based on the traditional Accelerated Robust Feature(SURF)matching algorithm to improve A SURF(speed up robust features)feature matching algorithm.By calculating the gradients in the X and Y directions of the left and right images,the texture information of the image is obtained,and an image containing edge texture features is obtained.Accurate feature points are used in accelerated robust features(SURF),and then the feature points are matched.Feature point matching pairs are finally used to complete interactive 3D reconstruction using 3D coordinates and Open GL texture mapping.The main work of this article is as follows:(1)The imaging model of the binocular camera is studied,and the parameters of the binocular camera are obtained using Zhang's calibration hair,which provides data for the subsequent calculation of three-dimensional space coordinates.(2)Study and analyze the Accelerated Robust Feature(SURF)matching algorithm.Based on the original algorithm,the Sobel edge detection operator is integrated to improve the SURF algorithm.First,the gradient information in the vertical and horizontal directions of the image is calculated to obtain Images with edge texture features,then calculate the image integration of the left and right images,construct the Hessian matrix,construct the SURF initial space,set the threshold for the Hessian determinant and determine whether it is a feature point,calculate the main direction,and extract the accurate feature points.(3)Pre-matching the feature points by the nearby fast search algorithm(Flann),and then introducing the random sampling consistency algorithm(RANSAC)to filter out the mismatched matching pairs,further improve the matching accuracy of the feature points,and establish the left and right images.Relationship.The 3D space coordinates of the target image are calculated by combining the data obtained by camera calibration,and preliminary 3D structure information is obtained.Then,a complete 3D image is obtained through triangulation and texture mapping.At the same time,in order to facilitate the observation of 3D effects,in this subject,An interactive function will be added to realize the human-computer interaction function of viewpoint rotation,and binocular interactive 3D reconstruction is realized.