Illumination invariant Chinese chessboard reconstruction based on color image

This paper introduces a new approach based on binarized Gabor filters and Hough transform to reconstruct Chinese chessboard by computation. Firstly, Gabor transform is performed directly on an original color image of chessboard to obtain Gabor filters in which only interested lines are kept by selecting appropriate parameters. Secondly, Gabor filters image is binarized and thinned and then lines are detected by Hough transform. Thirdly, a binarized Gabor filters´ image is affine rectified and three parameters which dominate reconstruction of chessboard are measured and coordinates of any crossing points in the chessboard can be calculated based on priori information and measured data. Finally, experiments have been conducted to verify illumination invariant feature of proposed approach. Results reveal that this new approach is more robust to illumination change. Comparing to conventional method, it is obviously superior and has been applied in a robust machine vision system for Chinese chess playing robot.