Research On Data Modeling Algrothim And Thechonology Of 3D Reconstruction

Computer vision is a comprehensive discipline whose research object is to make computer have the ability of understanding 3D environmental information from 2D images or views. This paper is focus on 3D reconstruction which is one of the most important in computer vision.Based on digital camera, a system scheme of 3D reconstruction is brought forward after characteristic of digital camera in this paper, and a 3D reconstruction system is realized and completed at last.First, an algorithm for camera calibration is examined in the paper. Based on characteristic of digital camera, interior and external parameters of digital camera are separately calibrated in this algorithm. The interior parameters can be calibrated in laboratory. Another calibration technique that is called right angle and three points is studied for digital camera external parameters. Calibration device is simple and only three points constitute it in this method which simplifies calibration device and process of calibration.Then this paper researches on stereo matching in 3D reconstruction. A new feature point detector for color image based on gray level Harris detector and Gauss-Laplace detector is proposed at first. Then the paper put forward a method which uses area-based algorithm, gradient of disparity constraint and homography matrix robust estimation synthetically to get the result of corners matching and make 3D reconstruction well done.Finally, 3D reconstruction realizes from Delaunay triangulation and texture mapping and can be displayed by OpenGL. It is proved that the effect of the system is satisfactory by actual application.