Research On Camera Calibration Based On Improved Genetic Algorithm

Vision is the main approach for people to apperceive the condition word and to know the outside word. With the development of computer technology and the increasing requirement for digital info, computer vision and binocular stereo vision develop greatly. Resumed from 2D image to 3D geometry structure (named 3D-Reconstruction) is the main purpose of computer vision. Camera calibration is the pivotal step in 3D-Reconstruction, and the premise and basement of obtaining 3-D space information by computer vision, also the main part of binocular stereo vision.The main work this paper is to research the technology and the method relation to the camera calibration, to provide reliable data and lay a good foundation for next step of computer vision and binocular stereo vision by obtaining the parameters of camera and striking up the relation between 3D space and 2D image,. This paper describes the achievements in binocular stereo vision at home and abroad in detail based on Marr's computer vision rationale. The focus is on researching camera calibration method and the early feature point detection process.First, research image feature point extraction issues by the numbers. In the process of corner detection, the elements, the methods and the capability of SUSAN operator, MIC operator and Harris operator are analyzed and contrasted, then using VC6.0 to complete different images' Harris corner detection. The detection results showed that the speed of Harris corner detection method was fast and this method can get even and accurate corners. So this algorithm was stable.Secondly, research camera calibration method and the related theory. The principle and the research status quo at home and abroad of calibration technology in stereo vision are described in detail. Then, the existing calibration technology and the difficulties in calibration technology at present are analyzed. In this paper, the Genetic Algorithm is used in camera calibration technology, and then analyzes the disadvantage and shortage in traditional Genetic Algorithm. Against these disadvantages and shortages, an improved method is put forward -adding a self-correcting searching interval and it conquers the traditional Genetic Algorithm's disadvantages when using in high-dimension space optimization. In crossover probability and mutation probability, adaptive crossover and mutation probability is adopted and it conquers the difficulties in convergence. The improved Genetic Algorithm has a greatly improvement in calculation accuracy and speed. At last, the method is applied to the experiment. By comparing the results, it showed that the method presented in this paper can partly improve the operation speed and calculation accuracy while simplifying experiment equipment and reducing costs and it proved the method's validity and feasibility.