Over exposure and camera lens distortion often result in the separation and local area asymmetry of a chessboard corner and the existing chessboard corner detection algorithms can not extract the corner information in such conditions accurately. Therefore, this paper proposed a detecting algorithm based on the circular template under an image coordinate. The symmetry and interchangeability needed by gray distribution of the local area for the chessboard corner were analyzed, and the properties of a convoluted image for the circular template were obtained . On the basis of the properties, the chessboard corner was defined and extracted. Finally, the symmetry of local redundant corner distribution was used to remove the redundant corner to improve the corner detection accuracy and to allowed the extracted accuracy of the corner to sub-pixel level in merely one step by employing the image coordinate. Experiment results show that our algorithm can achieve better results in over exposure and lens distortion both at simple backgrounds and complex scenes, and it is characterized by higher operation speed and smaller errors . Applying proposed algorithm to a camera calibration, a re-projection error less than 0. 3 pixels is obtained.%曝光过度和镜头畸变将分别导致棋盘格角点分离和角点局部区域不对称,现有的角点检测算法难以准确提取棋盘格角点.为此本文提出了一种图像坐标系下基于环形模板的棋盘格角点检测算法.该算法通过分析棋盘格角点附近的灰度分布应满足的对称性和灰度交替性等性质,得出环形模板卷积后的图像应满足的性质.利用该性质来定义并提取棋盘格角点,最后利用局部冗余角点分布的对称性来去除冗余角点,使角点检测更精确从而使提取的角点直接达到亚像素精度.实验结果表明:本文提出的棋盘格角点检测算法在曝光过度,镜头畸变和复杂背景情况下均能取得较好的棋盘格角点检测效果,且运算速度快,误差小.将该算法应用于实际摄像机标定,结果显示重投影误差在0.3个像素以内.
展开▼
