3d Reconstruction From Image Sequences

Vision is an important way for human to access to information and understand the world. Computer vision use computer to simulate the visual function of the human eye, extract information from the image or image sequences, and identify the shape and movement of the three dimensional scene and the objects. In the last three decades, there are rapid development in computer vision technology, and the scope of its application throughout the industrial, agricultural, medical, military and other fields.Three dimensional reconstruction is a technology to examine how to obtain three dimensional information through the two dimensional objects in space, and it has broad application prospects. It involves a wide range of disciplines, including the feature point extraction, image matching, multi-view geometric theory, numerical analysis, computer graphics, three dimensional modeling and other fields.This thesis discusses an algorithm to automatically compute a fundamental matrix between two images basing on the SIFT key-points and the RANSAC as a search engine. The input to the algorithm is simply the images, with no other priori information required; and the output is the estimated fundamental matrix together with a set of interest points in correspondence.This thesis designs and completes a three dimensional reconstruction system of scene from image sequences. A detailed description of the system of processes is discussed deeply, which relates to the feature point extraction and matching, the relationship between images, the updating of structure and motion, self-calibration, dense matching techniques and so on. Experiments show that in this thesis, the realization of the three-dimensional reconstruction system can be used to restore the appearance of the actual object and the reconstruction in high accuracy.