A Robust Algorithm for Go Image Recognition in Go Game

In this paper, image of go game recognition is proposed based on Hough transform algorithm. Firstly, preprocess the go image to reduce the amount of noise. Secondly, locate four corner coordinates by Hough transform to detect lines of unoccupied board for the opening of the game. Next, most significant difference from previous work is our use of simple geometric transformation model and in process of finding the grid. Another significant improvement is the use of Hough transform for circle around intersection to locate the piece. Finally, for each intersection, compute average brightness values around it and K-Means algorithm can be determined to quantize the colors. Experimental result in abundant go images shows that the proposed method has a good recognition performance on go images of different views, go boards and a certain degree of illumination nonuniformity.