Research On Camera Calibration And Stereo Matching Technology Based On Monocular Vision

As the theory of computer vision is becoming more and more perfect,the related technologies of stereo vision has been applied in such fields as vehicle navigation,object recognition,virtual or augmented reality and aerospace.As an important branch of computer vision,image based 3D reconstruction has deeply influenced people's lifestyle.Especially monocular vision based 3D reconstruction,it has become a research focus because of its simple structure and low cost.The main work of this paper is focused on camera calibration and stereo matching of feature points.In the part of calibration,Zhang Zhengyou calibration method is decided to use to calculate the internal and external parameters and the distortion coefficients of the camera,so as to correct the pictures taken by the camera.In this part,the main work is to extract and sort the detected corner points.Considering the characteristic of the corner points of checkerboard,a corner extraction algorithm based on Hessian matrix is proposed,including improve the method of corner recognition to achieve more accuracy.To accuracy the positioning of the corner points,a “Gray center of gravity method” based on improved weight is adopted.A new sorting algorithm of “four poles” is proposed,based that,the detected corner points of checkerboard are sorted.In the stereo matching part,considering the low matching accuracy rate of the original SIFT algorithm,a SIFT feature matching algorithm based on improved Harris algorithm is proposed.Using an improved Harris algorithm to extract corner points in image scale space,and a new corner scale factor assignment strategy is adopted to improve the accuracy of feature matching.Finally,through the experiment,the algorithms and improvements in the calibration part and the stereo matching part are verified,and the validity and rationality of the proposed algorithm are proved.