Analysis and Modeling of Two-CCD Camera Stereo Vision System

• Main Authors: Te-Tan Liao, 廖德潭
• Other Authors: Psang Dain Lin
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/25100988131552214759

博士 === 國立成功大學 === 機械工程學系碩博士班 === 91 === 　　The purpose of this paper is to analysis and modeling of two-CCD camera stereo vision system by utilizing the skew ray tracing method. This method is derived from a 4×4 homogeneous coordinate transformation matrix and the Snell’s law of the geometrical optics. 　　We first apply the revolution geometry, homogeneous coordinate transformation matrix and Snell’s law to develop the mathematical model of the direction of refracted (reflected) ray and the position of incident point of light rays while crossing medium boundary surfaces. Therefore, with this model, we are able to investigate three important topics of the geometrical optics: 1) The direction of a refracted (reflected) ray is determined according to Snell’s law. 2) Sensitivity analysis expresses differential changes of refracted (reflected) rays in terms of differential changes of incident rays. 3) Aberration of polychromatic light is presented analytically. Then the corner-cube retro-reflector is analyzed by the skew ray tracing methodology and the image orientation of Pechan prism is determined by sensitivity analysis. 　　An accurate binocular stereo vision system can greatly simplify solutions of many important vision problems such as the 3-D coordinate measurement, the mobile-robot visual guidance problem, and photogrammetry. However, precise modeling of binocular stereo vision systems is difficult because there is no fundamental imaging formation theory, and as a result, only the simplest pinhole model has previously been utilized traditionally. This paper has demonstrated that a binocular stereo vision system can be modeled by the skew ray tracing method to study the geometric relations between the binocular camera images and the 3-D surface. Numerical and experimental verification indicate that the performance of the proposed method is more excellent than the traditional pinhole model.