基于影像序列的相机分步自动标定方法

相机的自动标定是计算机视觉中根据影像序列进行场景欧氏重建的关键问题。在引入计算机视觉几何中若干基本概念的基础上,讨论了基于影像序列分步进行相机自动标定的基本原理,推导了该自动标定方法的实用计算公式,提出了采用牛顿迭代法进行参数答解的方法。同时,首次提出了一种三像标准投影矩阵组向同一参考系投影矩阵的线性转换算法。实验结果表明,所述的基于影像序列的相机自动标定方法是正确有效的。

Stratified Approach to Camera Auto-calibration Based on Image Sequences

Camera auto-calibration is one of the key problems of metric reconstruction based on image sequences in computer vision.On the basis of a few fundamental concepts in multiple view geometry in computer vision,the principle of stratified approach to auto-calibration was discussed in this paper.The corresponding practical algorithms of the proposed camera auto-calibration and solutions of parameters with Newton iteration were put forward.At the same time,a linear algorithm of transforming the independent canonical camera matrix triplets into canonical projection matrices in a global frame was put forward to avoid the onerous process of bundle adjustment in traditional projective reconstruction of image sequence.Experiments showed that the autocalibration method described in this paper was practical and efficient.