一种基于主动视觉的摄像机内参数自定标方法

AN ACTIVE VISION BASED CAMERA INTRINSIC PARAMETERS SELF-CALIBRATION TECHNIQUE

摄像机定标是计算机视觉的一项基本任务．本文提出了一种基于主动视觉的摄像机内参数自定标方法．该方法和文献［1］提出的方法相比最大的优点在于不要求摄像机在三维空间作任意方向的平移运动，只需要控制摄像机在同一平面作四组运动，其中每组包括两次相互正交的平移运动，同时适当调整摄像机平移运动的姿态，可以线性求解摄像机内参数．理论上证明了如果摄像机作三次俯仰或者作两次俯仰一次扫视来调整摄像机平移运动的姿态，则解存在而且唯一．通过稳定性分析证明前一种方法对误差很敏感，后者则是稳定的．最后给出了采用真实图像的实验结果．

Camera calibration is one of fundamental issues in computer vision. Inthis paper, a new camera intrinsic parameters self-calibration technique is pro-posed. Unlike the one presented by Ma Song-De in 1996, this approach does not re-quire camera 0f the ability to translate along arbitrary three dimensional directions.By controlling a pan-tilt-translation camera to translate along two orthogonal direc-tions four times, while regulating camera pose in fixed pattern, the camera intrinsicparameters can be solved linearly. It is proved that the solution is existing and u-nique if the camera pose is adjusted by either three tiIts or two tilts and one pan. Inaddition, based on simulations and experiments with real images, it is proved thatthe first one is numerically instable and quite sensitive to noise, whereas the secondone is robust.