Based On The Ccd Image Of The Three-dimensional Reconstruction Techniques

The traditional 3D modeling which are low efficient and high costs can not meet the needs of market in virtual reality. With the development of computer vision, 3D reconstruction which based on inverse project innovates continuously and promotes the application of 3D graphic technology in the relevant field.On the basis of researching the 3D reconstruction problems which include calibration error, matching precision and real-time reconstruction. The paper realizes a mutual 3D reconstruction system. We use the method presented by Tsai to calibrate the single camera and rectify the lens distortion. Furthermore, we do the double cameras calibration and epipolar rectification. Then we use the median filter to eliminate the noise in twain images and use an appropriate matching strategy to get the disparity map. The paper presents an approach that getting the triangulation of 3D scattered points directly. We use the growing method to extend the three edges of each triangle until all the triangles have been finished. Finally, we construct the 3D model by OpenGL in Visual C++ 6.0.For the purpose of reconstructing the object's surface accurately, we present a new image matching scheme which use the position information of the relevant point to construct an energy function, and minimize the energy by using partial differential equation. The energy function is composed of regularization term and matching term. The matching term reflects the disparity of the relevant point, the regularization term reflects the smoothness of the image. The method has high precision and the disparity map is dense. The achievement of the method is very important for the 3D reconstruction technology.