1.1 Objectives

• The system must capture an image of the board before and after a human move.
• The system must find the edges of the chess board.
• The system must locate individual squares within the board.
• The system must determine where the board has changed, thus indicating a move.
• Communication must be made with the chess logic system to send the move.

1.2 System Overview

Ultimately, the job of the vision system is to let the chess logic system know when a move has been made.

The system itself is comprised of a SGI camera attached to an SGI O₂ workstation which captured the pictures, as well a another Linux box which runs the image processing software we developed to analyze the pictures.  

The camera is a very basic model, not at all intended for vision, so it was the cause for some of the limitations we encountered.  The camera is mounted from an adjustable metal arm a few feet above the chess board.  The chess board we made ourselves.  We used red squares, because red provided a high contrast with the color of the pieces.

1.3 Review of Similar Projects

There are not a lot of similar projects available for study.  Certainly there are several chess playing computers, and even a few chess playing robots, but most of these robots have simply been concerned with moving the pieces, hence vision has not been a part of most of these projects.

The closest thing to the vision system used for Machbanai is a Robot Computer Chess Game developed and patented by some computer scientists at the California R&D center.  However, although their system also uses a form of vision to determine a human's move, the inventors used a complicated magnetic system, and not a camera as is the case with Machbanai.

We drew from only basic concept in computer vision and did not have any examples of chess robot vision systems to examine in the development of Machbanai.