基于单平面模板的摄像机定标研究

Research on Camera Calibration Using a Planar Pattern

提出了一种摄像机定标方法 ,只需要摄像机从不同方向拍摄平面模板的多幅图像 ,摄像机与平面模板间可以自由地移动 ,运动的参数无需已知。对于每个视点获得图像 ,提取图像上的网格角点 ;平面模板与图像间的网格角点对应关系 ,确定了单应性矩阵 ;对每幅图像 ,就可确定一个单应性矩阵 ,这样就能够进行摄像机定标。该算法先有一个线性解法 ,然后基于极大似然准则对线性结果进行非线性优化求精。该方法同时也考虑了镜头畸变的影响。实验结果表明该算法简单易用。

A camera calibration using a planar pattern is proposed. The planar pattern can be freely moved, and the motion parameters need not to be known. The planar pattern consists of a grid pattern imaged from different points of view. For each view, corner points are extracted to calculate the correspondence between the image plane and the planar object. The correspondence is in the form of a homography matrix. Then for each view, a homography is established for camera calibration. The procedure consists of a closed-form solution followed by a nonlinear refinement based on the maximum likelihood criterion. Radial and tangential len distortion is mo-deled. Real data can be used to test the technique, and results demonstrate that the technique is easy to be used.