Browse Prior Art Database

Method for using a webcam to automatically read moves played on a chess (or checker) board

IP.com Disclosure Number: IPCOM000128875D
Original Publication Date: 2005-Sep-20
Included in the Prior Art Database: 2005-Sep-20
Document File: 2 page(s) / 46K

Publishing Venue

IBM

Abstract

A method for using a webcam or other digital image capturing system to input moves for board games, such as chess. The most interesting point of the proposed solution is that it does not require precise image recognition. In fact it doesn't have to recognize anything, as long as it is able to tell *where* the two images differ. This information, integrated with knowledge about the board geometry and the rules of the game, allows the software to understand which piece moved, and where. This is usually much cheaper than implementing a sophisticated recognition system. Moreover, this approach is not limited to board games and it can be easily applied to other fields as well.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 2

Method for using a webcam to automatically read moves played on a chess (or checker) board

Main Idea

The solution allows a computer to read moves played on a chessboard when playing for example chess or checkers, by using a webcam pointing to said chessboard.

     All solutions I know of this problem are very different from this one, and go from using sensors (e.g. magnetical or pressure sensors) to tagging the piece bases with easy to recognize optical marks and playing on a surface that allows to read such marks, e.g. a scanner.

     The proposed solution works by elaborating the images that come from the webcam. However, the approach used does not involve complex image recognition, which would be very difficult to do, but rather uses knowledge about the initial status of the board and the rules of the game to greatly simplify the image recognition task, so that it becomes actually manageable with a standard webcam and a PC.

In summary, it works as follows:
1) the camera is allowed to take a shot at an empty board, so that the software can "understand" the board position and (pre)compute the necessary transformations: this is easy to do as for our purposes the board is a bidimensional object with a simple two-color pattern on it; 1b) the need for step 1 actually depends on: the quality of the software image recognition and the type of pieces placed on the board. For example, checkers are usually small and "flat" and won't interfere much with the recognition process;
2) pieces can now be placed on the board, and the piece placement must be communicated to the software: this allows the software to see how the piece "perturb" the original board pattern, e.g. by covering some extra squares due to perspective and so on; 2b) note that in the vast majority of cases it's easy to setup the board position in software, as it will be either the standard initial position or it can be obtained by moving back and forth into a known list (such as browsing a game);
3) when a move is being played, there will be something very "confusing" going on the board, such as for exam...