Go Chess Game Image Recognition On Mobile Handheld Terminal

The research of Go Chess Game Image Recognition on Mobile Handheld Terminal is aimed at determining the result of a go chess game automatically. Traditionally in the end of a go chess game, chess pieces are manually counted to decide the winner and the loser. This not only consumes much time, but could also introduce mistakes. In this paper we propose a go chess chessboard recognition algorithm based on projection, and port the algorithm to Android platform. Thus game result can be automatically determined with the captured image of the chessboard.The essence of chess piece recognition algorithm based on projection is geometric transformation of the chessboard image, the segmentation of black and white chess pieces and the projection of image. First of all the purpose of geometric transformation is to eliminate anything else beside the chessboard. Only after correct geometric transformation, which is the preliminary step, can follow-ups be done.Then image segmentation of chess pieces is achieved by histogram threshold method. Black pieces are segmented using grey images; on the other hand, white pieces are extracted by a proposed algorithm which segments the image according to the difference between B (Blue) components in RGB vectors of white and yellow. Experiments show that the algorithm can segment the chess pieces precisely.In addition, to recognize the lines on the chessboard, a line detection algorithm based on projection is proposed. First, apply opening operation to eliminate isolated noises. Second, add the black and the white image, and project the added image along the horizontal and vertical axis, respectively. Finally, locate the lines according to values of projection. Experiments prove the algorithm can detect the chessboard lines precisely. After that, detect the location of black and white chess pieces on the basis of detected chessboard lines. Then, fill in the vacant cells according to chess piece filling rules. Finally, count the number of black and white pieces respectively to decide the winner/loser.Finally, the algorithm is ported on Android platform. Thus, instead of manual counting, go chess fans or referees can conveniently determine game results with their mobile phones or pads.