利用二次曲线拟合和圆环点进行摄像机标定

为了提高摄像机标定的精度和实用性，提出一种新的标定算法．所用的标定模板由平面上两个相交的圆周组成，且两圆的圆心和半径均未知，通过对两圆的图像进行二次曲线拟合，再根据拟合的二次曲线来计算圆环点图像完成标定过程．与经典的平面标定算法相比，该算法无需进行角点检测和角点匹配，不需人工干预即可实现自动化标定，且在标定过程引入了更多的图像点信息．模拟实验结果表明，在实际的噪声水平下，该算法对焦距的相对标定误差比平面标定算法减小约0．1%，此时主点的相对标定误差也略小于平面标定算法．真实图像的实验结果也表明，所提算法具有较好的精度和实用性．

Camera Calibration with Conics Fitting and Circular Points

To improve the accuracy and practicality of camera calibration,a novel algorithm is proposed by using a new calibration model which consists of two planar intersection circles.The radii and the centers of the circles are both unknown.The algorithm uses conics fitting to calculate the images of the circular points from the images of the circles.Compared with the plane based algorithm,it can avoid the corner extraction and correspondence problems,and automatic calibration can be carried out easily without human interference.Besides,it uses more image points for calibration.The simulation results show that the relative error for the focal length is reduced by 0.1% compared to that of the plane based algorithm under the realistic noise level(0.5 pixel).And the relative error for the principal point is a little smaller as well.The results with real image data also demonstrate that the proposed algorithm is accurate and practical.