One of the main components of the computer vision is 3D reconstruction. Calibration of camera' s interior and external parameters, feature selection and stereo matching are key techniques of 3D reconstruction of the computer vision .In the computer vision, the input image is the 2D image' s gray which is a function included many factors such as the geometry character of 3D object, lighting, the surface character of material, the color of objects and the parameters of camera . The process of Computing these parameters by gray is a inverse problem which is almost non-linear. The result of the problem, which is sensitive to the errors arising from noise and discreteness, is not unique. So the computer vision itself is ill-posed and the difficulty of 3D reconstruction is how to get the robust result.A system of 3D reconstruction is proposed and implemented in this paper, which is based on the theory of binocular stereo vision. With the system, user can calibrate the camera with Zhang' s flexible new method, select feature points and get a robust matching points by initial correlation matching, medium robust matching and strong robust matching. Finally user can realize the structure reconstruction from motion and display it using OpenGL. It is proved by the system that the method is robust and the effect is satisfactory.
|