Research On 3D Reconstruction Technique Based On Uncalibrated Camera Binocular Vision

The 3D reconstruction based on binocular vision creates the 3D model of the scene according to binocular vision measurement principle,by using the parallax of the two images,which is obtained by the two cameras in different locations or one camera rotated and translated shooting the same scene.Because of the low cost of image acquisition equipment and the wide application of refactoring technology,the 3D reconstruction technology based on binocular stereo vision has become an important research method of the 3D reconstruction techniques.How to calibrate the camera correctly,improve the accuracy of the stereo matching,estimate the fundamental matrix and solve the 3D coordinate of the space accurately is the key point and difficult point in the study.In this paper,the above problems are studied.Firstly,the camera's calibration method is described in detail.For the standard test image,the camera calibration method based on the vanishing point is used,the Hough transform can be used to realize the camera calibration without knowing the camera's motion attitude and related parameters.For the self-shooting images,Zhang's plane template calibration method is used,the use of calibration template improves the calibration accuracy.Secondly,the performance of SIFT algorithm and SURF algorithm is compared,the SIFT algorithm is used as the research foundation of the follow-up work in consideration of correct matching rate,and an improved method of error matching is proposed.Before the fundamental matrix is obtained,some obvious false matching points can be eliminated,which can improve the precision of the basic matrix.Then,an improved fundamental matrix estimation method is proposed.This method combines the advantages of RANSAC algorithm and M-Estimator estimation algorithm,and can estimate the fundamental matrix more accurately.The matching points are restrained by the estimation results,then the mismatching points are eliminated again so as to improve the matching accuracy of feature points.Finally,the motion recovery method is used to recover the rotation matrix and the translation vector of the camera by taking part in the fundamental matrix and internal parameters of camera.For the standard test image,the spatial coordinate of the space is solved,eliminate the matching points seriously deviated from the main scene according to scene depth uniformity.In order to express the geometry information of the scene more clearly,triangulation of the spatial 3D points is performed.In order to restore the realistic 3D scene,texture mapping is realized by using Open GL in the VS2010 development environment.In order to verify the versatility of the algorithm,the above process is completed for the self-shooting images,Effective recovery of self-shooting scenes is achieved.The results show that the three-dimensional reconstruction algorithm is reliable.