Robust Recognition of Checkerboard Pattern for Deformable Surface Matching in Multiple Views

Checkerboard pattern can be used in many computer vision areas by matching the pattern as a surface, such as camera calibration, stereo vision, projector-camera system and even surface reconstruction. However, most existing checkerboard pattern recognition methods only work in planar and fine illuminating circumstances. A robust recognition method for checkerboard pattern is proposed in this paper to deal with those arbitrary surface deformation and complex illumination problems. Checkerboard internal corners are defined as special conjunction points of four alternating dark and bright regions. A candidate corner’s neighbor points within a rectangular or a circular window are treated as in different onepoint-width layers. By processing the points layer by layer, we transform the 2D points distribution into 1D to detect corners, which simplifies the regions amount counting and also improves the robustness. After corner detection, the pre-known checkerboard grids rows and columns amounts are used to match and decide the right checkerboard corners. Regions boundary data produced during the corner detection also assist the matching process. We compare our method with existing corner detection methods, such as Harris, SUSAN, FAST and also with the widely adopted checkerboard pattern recognition method, FindChessboardCorners function in OpenCV, to show the robustness and effectiveness of our approach.