一种鲁棒的高精度摄像机标定方法

Robust accurate camera calibration method

为了解决高精度摄像机标定过程中标定块制作难度大、标定方法对实验设备要求高、实验结果噪声鲁棒性差等问题,提出了一种新的标定模板及配套的标定算法。标定模板用十字网格代替经典的棋盘格,在强光照和强噪声环境下能够得到更精确的位置信息;标定时利用灰度加权求和的方法提取标定图像中心点,以该中心点为基准设置角点的匹配模板,通过计算相关程度确定其余角点坐标,并进一步确定精确坐标,最后利用BP神经网络来隐性标定摄像机参数。实验结果表明,该方法在强光照、强噪声的环境下也能获得高精度的标定结果,满足精密测量系统的自动化标定应用。

In order to solve the problem of the difficulty of making calibration block, the high requirements on the experimental equipment, the bad robustness against noise, this paper proposed a new calibration template and a calibration algorithm. The new template used cross grid instead of the classical checkerboard, thus got more accurate position information in strong light and noisy environment. This algorithm used the method for gray weighted summation to extract the center point of image, based on which the matching template was set. It determined the coordinates of other points by the related degree, and further determined precise coordinates. Finally, the method adopted BP neural network to calibrate camera parameters implicitly. The results indicate that the method can obtain precise calibration results even in strong light and noisy environment, which meets the requirement of automatic calibration applications in precision measurement system.