Research On Camera Calibration Methods

With the development of computer science, obtained image information by computers could hardly satisfy with us, thus a novel comprehensive branch named computer vision is come into being. Three-dimensional reconstruction as an important part of the computer vision is aimed at reconstructing geometry of the three-dimensional object from the two-dimensional image. Building the relationship of the transformation between two-dimensional images and three-dimensional object is the so called camera calibration.This paper gives brief summary on fundamental theories of camera calibration. Our works are mainly on realization of a calibration algorithm based on a hypothesis that after correction the straight line should be straight either. Focused on those contents on camera calibration, the research achievements are listed as follow:1. Making a description on fundamental theories of camera calibration. Sub pixel corner detection was realized by Harris algorithm. Edge extraction of2-D image was also realized by Canny algorithm.2. Giving a description on conventional camera calibration and the realization processes of those algorithms. Applicable situations and merits of those algorithms are compared based on precision and operability. Zhang Zhengyou calibration algorithm was detailed analyzed on the modeling, solution procedure and its merits through the realization of such algorithm.3. The straight line constraint based non-measuring calibration algorithm was proposed based on a hypothesis that after correction the straight line should be straight either. The procedure of such algorithm is that firstly the modified line is estimated by linear fitting and then the distortion parameters are optimized by Levenberg-Marquardt non-linear least square method.4. A novel and applicative calibration target for straight line constraint based non-measuring calibration algorithm were proposed and designed. Such calibration target was proved to be superior to the common calibration target. Calibrations on different situations were realized based on such algorithm by using Matlab programming language. Its superiority and boundedness were demonstrated when compared with Zhang Zhengyou calibration algorithm. The inapplicable situation of such algorithm was also proposed and stimulated. The sensibility of the initial value was also evaluated.