基于形状识别的X型棋盘格角点检测

X Corner Detection in Chessboard Image Based on Shape Recognition

中文关键词:  X型角点  形状识别  极坐标  特征选择

英文关键词:X corners  shape recognition  polar coordinates  feature selection

中文摘要:

      棋盘格角点检测是相机标定的一个重要环节，其准确率直接影响相机标定的精度，常规角点检测存在对棋盘格图像质量要求较高、检测准确率不高的问题。对此，依据棋盘格图像灰度分布的对称性，利用双层增强滤波器将其角点区域增强为X形状，待检测角点即为X形状中两个分支的交点，将角点检测问题转换为X形状的识别。进行X形状识别时，首先提取待识别区域像素的极坐标作为形状识别的初始特征集合，经特征选择后依据角度特征进行X形状识别。最后在识别出的角点区域内通过直线拟合确定亚像素级角点。实验结果表明，所提出方法进行棋盘格角点检测的准确率较高，且抗干扰能力较强。

英文摘要:

      Chessboard corner detection is an important part of camera calibration,and its precision affects calibration accuracy.The conventional corner detection methods have the problems of high image quality requirement and low detection accuracy.Based on the symmetry in gray level distribution of check board image,the corner region is enhanced to X shape by a double-layer filter,and the corner point is found at the intersection of two branches of X shape,thus corner detection is converted to X shape recognition.When X shape is recognized,the polar coordinates are selected as initial feature set,then,after feature selection,the X shape is recognized by the angle information.At last,sub-pixel-level corner points are detected by line fitting in corner region.Experimental results show that the proposed method is more accurate and has better anti-interface ability.