Researches On Camera Self-calibration And Stratified Reconstruction

In this paper, we mainly discuss several problems in the field of 3-D computer vision, which are the algorithms for based on vanishing points to calculate the other geometry structure relationship, the Self-Calibration and the stratified reconstruction, etc, Our main work is as following:(1)Based on vanishing points to calculate geometry structure relationship. In the appliance of object identify and scene analysis, we need comprehend the geometry structure relationship. If camera is not calibration how to use the scene's information and vanishing points to accomplish it. Then we give two algorithms and introduce the range of apply. The end, we give a experiment. The result indicates it is efficient.(2) The Self-Calibration. The algorithm of camera's Self-Calibration is always a important research domain. In this paper by taking advantage of parallel lines and orthogonal lines in architecture as usual, we can calculate the absolute conic image and vanishing points. By matrix decompose we at last obtain the camera calibration matrix. At the same time we analyze how to calculate the camera calibration matrix based on 3 parameter model in one image. We advance some conclusion also. Simulation experiment compare two results. The result indicates it is efficient by reconstruction of scene.(3)The stratified reconstruction. In this paper we discuss if two cameras have the same calibration matrix how to use some knowledge of scene's to stratified reconstruction from a pair of views .the key to the projective reconstruction is compute the fundamental matrix. The essence of affine reconstruction is to compute the infinite homographyo The key to metric reconstruction is identification of the absolute conic image, in the end we give a experiment. The results indicates it is efficient.