Chessboard corner detection under image physical coordinate

Abstract

The existing chessboard corner detection algorithms cannot be used in complex scenes because distortion might occur as a result of overexposure or low resolution, among other factors. This distortion hinders the precise detection of the chessboard corners using the previous methods. To address this issue, we proposed a chessboard corner detector based on image physical coordinates and a round template. The physical coordinates allowed our detected chessboard corners to reach the subpixel-level after only one step. We first covered the distorted chessboard corners by utilising the morphological dilation. Then, we employed our round template to pass through the dilated image and ultimately determine the chessboard corner coordinate by analysing the grey distribution of the traversed round template and calculating the centroids of redundant points. The experimental results showed that our algorithm performs better than other algorithms in both simple backgrounds and complex scenes. By applying our detector to camera calibration, we obtained a smaller re-projection error, thereby proving the validity of our proposed detector.