数字摄影测量系统中非量测CCD相机标定算法

Non-metric CCD camera calibration algorithm in digital photogram metry system

为了解决传统的相机标定方法,通常需要建立复杂三维标定块或高精度三维控制场这一实际问题,采用平面控制格网作为标定块,利用二维直接线性变换分解出相机的内外方位元素初值,然后采用改进的Hough变换算法检测标定图像中的格网直线并利用最小二乘法拟合出最佳直线,通过求直线的交点得到标定格网点的像坐标。最后利用自检校光线束法平差进行相机的精确标定。实际图像数据实验结果表明:主点和焦距的标定精度分别达到了0.2和0.3像素左右。可以满足高精度近景三维量测的要求。

Traditional camera calibration method usually need complicated 3D calibration block or high-precision 3D control fields. In order to avoid this tedious work, we propose a camera calibration method using planar control grid. First, camera＇s initial values of intrinsic parameters are determined using 2D direct linear transformation and nonlinearity equations. Image coordinates of planar grid control points are detected using improved Hough transformation algorithm and least squares line fitting. Finally, camera calibration with high accuracy is addressed using bundle adjustment with self-calibration. Good results have been obtained with real image data calibration and demonstrate the method we proposed in this paper is feasible. Calibration precision of principle point and focal length is about 0.2 and 0.3 pixels respectively, which can meet the requirements of close-range photogram me try with high accuracy.