Research Of Self-Calibration In The Computer Vision System

With the fast development of the intelligence and automation in the measurement system, the auto self calibration become a trend. Former calibration method applied either an outer high precise target or used more than one picture so they need manual help and is low efficient and precise. In this essay, we bring a self calibration method for the 3D vision measurement system. We built in a high precise target and another optical path. As a result, we can calibrate the system real-time and automatically.In this paper, the computer vision system and the classic calibration arithmetic are presented firstly. Then the introduction and the advantages of self-calibration arithmetic are proposed. Also, the self-calibration arithmetic is classified. In contrast to the former calibration methods, the self-calibration is much more agile so it doesn’t have lots of constraint as the classic ones in practical use. After a presentation of the classic camera model with relative basic knowledge, the paper discusses some of the useful and general camera self-calibration method, followed by the analysis of its principles as well as results.According to the needs of our subject, a newly optical and hardware design is proposed base on the built-in high precise target. After careful comparison of different kinds of target image such like rectangle, lattice, check board, parallel line and concentric circles as well as different kinds of image processing arithmetic such as spot detection, line detection, edge detection and round detection, the concentric circles images with round detection were selected and put in to further tests about their robusticity, repeatability, precision and distortion revision. An optimization method was followed. The result is well under theoretically analysis and experiments. The whole system was applied on LY404 multi-functional 3D computer vision measurement system which is a co-operate subject of Shanghai Jiao Tong University and Guilin Guang Lu corporation. The sample machine was made and had been exhibited in CIME in Oct 2008.At last, this paper made a conclusion about existed self-calibration methods and made a prospect of its future.