【摘要】 在计算机视觉的研究中 ,定标、运动、结构与匹配是常用到的 4个方面的信息 ,而定标是进行运动和结构分析的必要条件 ,不需要参照物而利用图像系列进行摄像机自定标问题在 2 0世纪 90年代初就有所突破。然而 ,大多数自定标研究仍然集中于在视觉任务完成过程中摄像机内部参数固定不变的情况 ,在内部参数发生变化时 ,这些方法还不能适用。扩展以前利用Kruppa方程进行摄像机自定标的方法 ,通过对基础矩阵进行奇异值分解(SVD) ,将Kruppa方程简化 ,得出了一种可处理内部参数发生变化时的摄像机自定标方法。仿真实验证明了该方法的可行性。更多还原

【Abstract】 Calibration, motion, structure and matching are the information often used in computer visual research. Estimation of the camera intrinsic calibration parameters is a prerequisite for motion and structure analysis. With the aid of image sequences instead of a calibration pattern, the introduction of the self-calibration paradigm in the early nineties is a major breakthrough related to the self-calibration problem. Until recently, however, most research efforts have been focused on applying the self-calibration paradigm to estimating constant intrinsic calibration parameters in visual tasks. Therefore, such approaches are inapplicable to the case where the intrinsic parameters undergo continuous changes. The previous method for self-calibration by use of Kruppa equations is extended. The method is based on the singular value decomposition (SVD) of the fundamental matrix, which leads to a simple form of the Kruppa equations. A camera calibration method is obtained to deal with the case where the intrinsic parameters undergo continuous changes. The simulation test demonstrates the feasibility of the approach.更多还原