3D Reconstruction Algorithm Based On Multi-views And Software Implementation

3D reconstruction from multi-views is one of the difficult problems in Computer Vision. During the last decade, it has become one of the most popular topics because of the appearance of multiple view geometry theory. 3D reconstruction is widely applied in pattern recognition, robot navigation, military affairs and so on. Moreover, it has also been well applied in the newly arising fields, such as virtual reality, image-based modeling and rendering, digit museum. Still, it will be one of the most important research fields in the future.This thesis presents a systematic research on 3D reconstruction based on multiple view geometry theory. Starting from camera model, on the basis of the research in image matching and camera calibration, the structure of the object can be computed using triangle method. Then the vivid 3D result can be displayed with the help of computer graphic method. In the cooperation with East China Photoelectricity Institute, with the aid of OpenCV, IPL, OpenGL, 3D reconstruction software was designed by using VC++.Some key problems in 3D reconstruction are addressed in this thesis. The main research works and achievements are outlined as follows.1. Image matching is studied at first. Image matching is a crucial technique in computer vision, motion estimation and pattern recognition etc, and it is also the first step to implement 3D reconstruction. It aims to find the best matches between two or more point sets when there are geometric distortions, point measurement errors and contamination present. On the basis of some classical constraints in image matching, an image matching algorithm based on multiresolution and epipolar constraint is presented in this thesis.2. The approach to camera calibration is investigated. Camera calibration is an essential step to obtain the information of 3D Euclidean structure form 2D images in the field of computer vision. The technique of direct calibration for intrinsic...