提出了基于直线模型的相机在线标定算法。该方法通过自动跟踪视频中物件的边缘, 在线求解相机系统内外参数, 同时建立物件边缘同图像边缘的对应关系。然后, 利用物件边缘同图像边缘的对应关系, 在相机内外参数初值的基础上, 通过构建融合边缘端点信息的误差函数, 迭代优化求解相机内外参数。进行了仿真和实物实验, 分别使用提出的基于直线的标定方法和传统的基于棋盘格内角点的方法标定了相机内外参数, 并对两种方法的标定结果做了对比。结果显示: 在仅使用边缘信息的条件下, 本文方法可以获得同传统的标定法一致的精度, 标定后重投影误差(RMS)为0.6 pixel。本文方法利用具有标准尺寸的物件即能实现相机内外参数的估计, 且无需制备平面靶板便可获得与传统方法相同的标定精度, 标定过程更为灵活, 有实用价值。

A camera in-line calibration method based on a line model was proposed. By automatically tracking the edge of an object, the internal and external parameters of the camera were solved linearly while the corresponding relationship between object edge and image edge was established. On the basis of the relationship and the initial values of internal and external parameters, the camera parameters were estimated optimally by developing a new error model between the model line and image line with the endpoint information. The simulation and experiment were performed. The internal and external parameters of the camera were calibrated by proposed method and traditional checkerboard interior angle method, and the calibration results were compared. The results only by using edge information show that the proposed method have the almost the same accuracy with the traditional method and the reprojection error(RMS) is 0.6 pixel. In conclusion, the proposed calibration method can implement the estimation of internal and external parameters only by using the object with a standard size and can obtain the same calibration precision as that of traditional method without preparation of the plane calibration object. It is more flexible and practical.