This paper presents a method to recognize Go stones on a Go board for the automatic records of Go games, which is robust to camera’s position and orientation. The algorithm first performs camera calibration to predict the geometric relationship betw...
This paper presents a method to recognize Go stones on a Go board for the automatic records of Go games, which is robust to camera’s position and orientation. The algorithm first performs camera calibration to predict the geometric relationship between a camera and a Go board. Then, it briefly renders the shape of a Go stone at each intersection of grids by analyzing rays from the camera center to an area surrounding of the intersection. According to the brief shape of the rendered stone, the method extracts a rectangular patch of the stone from a captured image. For safety, the size of the patch is less than that of the stone. Finally, our method decides whether the patch is from black, white, and no stones.