International Symposium on Nonlinear Theory and its Applications


Session Number:1-2-4



Accurate Extraction of Go-coordinates using Genetic Algorithm for Automatic Generation of Go-record

Kojiro SHIBA,  Shogo NISHI,  Kunihiko MORI,  


Publication Date:2005/10/18

Online ISSN:2188-5079


PDF download (437.7KB)

We develop the system which automatically carries out Go-record from the image of the Go-game situation. For the Go-record, Go-stone has to decide its position on the 19x19 cells of the Go-board which we call Gocoordinates. In our system, the Go-board is shot in various directions with the CCD camera. In this case, it is difficult to directly extract accurate Go-coordinates in the image. However, we are able to extract accurate Go-coordinates if the area of the Go-board is recognized accurately. But, it is difficult to manually decide the four corners of a Goboard in the appropriate time and at the pixel unit. Consequently, in the first step, the contour of the Go-board has to be automatically recognized from the image which includes unrelated objects. In this paper, we propose the pattern matching algorithm for the extraction of the accurate Go-coordinates in the image using the genetic algorithm (GA) within several 10 seconds. We evaluated the proposed algorithm on the prototype using the proposed algorithm by dealing with the image which includes unrelated objects. As a result, it has been verified that the prototype generates the correct Go-record.