Research On 3D Reconstruction Technology Based On Sift Feature Matching

Binocular stereo vision is an important research direction of computer vision,and it has become a new research hotspot because of its partial characteristics of human eyes.Its application in various fields is more and more.In this paper,binocular stereoscopic navigation of mobile robot System as the background,the image preprocessing,stereo matching and three-dimensional reconstruction technology were studied.In this paper,the data collected by binocular camera are analyzed and processed.Firstly,the adverse effects in the process of data acquisition are removed by preprocessing of camera calibration and image correction.The corresponding matching points in the left and right image pairs are solved by stereo matching based on Sift feature.The regional growth algorithm is introduced to obtain the dense matching point data.After the matching point set is selected by RANSAC method,the matching point with high accuracy is obtained,and the parallax image of the image pair is obtained and the color disparity map is introduced.The feature information and position information of the object in the image are described.Finally,according to the principle of binocular vision,calculate the three-dimensional coordinates of the point cloud.The noise points and the distortion points generated during the image acquisition are removed by the bilateral filter.Then,the edge points in the scattered cloud information are extracted by the improved point cloud reduction algorithm based on the preserved edge points of the octree.For the non-edge points Its normal vector deviation to determine whether to retain;to be refined after the point cloud data.After processing the cloud data is still scattered point cloud data,visibility is not high,this paper will be processed after the point cloud data Delaunay triangulation,to enhance its visibility.In order to improve the complexity of Delaunay segmentation and the shortcomings of traditional point-by-point insertion method,the 3D point cloud is projected onto the two-dimensional parameter plane,which reduces the computational difficulty of the algorithm.The point insertion method is improved for grid insertion,reducing the effect of the insertion point on the surrounding triangle.This paper studies the three-dimensional reconstruction by constructing the experimental platform,validates the accuracy and real-time of the algorithm,and analyzes the applicable scope and shortcomings of the algorithm.Which laid the foundation for the complete navigation navigation of mobile robot.