一种新的基于直线特征的摄像机自标定方法

A Novel Approach for Camera Self-calibration Based on Parallel Lines

通过引入场景的几何约束,提出了一种新的基于直线的摄像机自标定方法。利用射影几何中平行直线投影的交点与光心的连线平行于该平行直线的性质,给出了摄像机内参数的线性求解方法。只假设场景中存在垂直的平行直线,相对于文献[1]而言,其适用的情况更为广泛。实验结果验证了方法的有效性。

Camera calibration has been attracting a lot of attentions in computer vision field these days, for it is a necessary step when recovering 3D shape from its 2D images. All methods for calibration can be classified into traditional ones or selfcalibration ones. The main difference between them is that traditional methods need known calibration objects while self-calibration ones do not. The advantage of self-calibration makes a number of researchers studied on it. Hundreds of approaches have been proposed each year. However, most of these methods failed to get the parameters we want simply or failed to keep robust. The approach we proposed here enjoys not only simplicity but also robustness as well, for the reason that we induce the scene information to do calibration. The feature of vanish point of parallel lines has been well used, and experiment results showed that our approach is feasible and effective.