Corner Detection Method of Chess Board Based on Symmetry Analysis

A symmetry analysis-based corner detection method is proposed to realize automatic corner detection of a chess board for camera calibration. The method includes feature points extraction, initial corner detection, and corner detection. Firstly, the candidate feature points are obtained based on the relationship between corners and the image binary edge. The feature points of the chess board are acquired by clustering the candidate feature points and computing the symmetry for every cluster using the ChEss(Chess-board Extraction by Subtraction and Summation) detector. Then the concept of initial corner is presented, and the detected initial corners are set as the starting point for corner searching. The corner detection methods are proposed by considering their different types of symmetry. Finally, the sub-pixel corner location is estimated using the detected corner location as the initial value. Experimental results demonstrate that the corner reprojection precision is approximately 0.12 pixels. The proposed corner detection method can realize the automatic corner detection for a chess board, and satisfies the requirement for camera calibration.