摄像机标定的棋盘格亚像素角点检测

针对传统最小核值相似区(small univalue segment assimilating nucleus,SUSAN)算子对棋盘格角点检测时会将角点与边缘混淆的缺点,提出了一种运算速度快、定位精度高的改进SUSAN亚像素角点检测算法。根据SUSAN模板中角点的特性,提出了灰度对称度的概念,利用灰度对称度区别出棋盘格的角点和边缘。在利用改进的SUSAN算子进行亚像素角点检测时,综合应用了索贝尔边缘算子、灰度平方重心法等方法。提出的亚像素级角点检测方法快速有效,在摄像机标定试验中,其重投影平均误差小于0.2个像素。

Sub-pixel corner detection of tessellated picture based on camera calibration

The traditional small univalue segment assimilating nucleus(SUSAN) operator may confuse the corner and the edge.An improved SUSAN sub-pixel operator with high speed and high position accuracy was proposed.According to the characteristics of the SUSAN template,the gray symmetry degree was defined to distinguish the corner from the edge.Sobel operator and gray square centrobaric arithmetic were synthesized with the improved SUSAN operator in subpixel corner detection.In the camera calibration experiments,this method proved to be rapid and efficient,and the improved operator's average re-projection error was less than 0.2 pixels.