Electronic Abalone Referee

1
Electronic Abalone Referee
By Benjamin Bayes (EE, 2005) Dean Miller (EE,
2005)Advisor Prof. Rudko

• Enjoy the full gaming pleasure without dispute

2
In this presentation

• Introduction to the game Abalone
• Different options that could have been used for
  image acquisition for this project
• Various MATLAB commands used to process the image
• A demonstration of the game and of the designed
  referee system

3
Introduction

• Abalone is a board game involving marbles that is
  based heavily on strategy.
• The object is to push 6 of your opponents off of
  the board before it happens to you

4
Rules of The Game

• Starting Position

• Broadside Move

• In Line Move

5
Rules of The Game

• Sumito (Pushing)

• Impossible Moves

• Ejecting

6
Rule Variations

• Alternate Starting Positions

• Rules for 3-6 Players

7
Project Objective

• To make a digital referee for an unaltered
  Abalone
• Referee would sense every move made
• Would identify legal moves, and flag illegal ones
• Allow the players to reset the game if move is
  illegal
• Would display score, moves, and a digital
  representation of the board
• Compatible with 2 players (due to time
  restraints, we were unable to expand it to a
  possible 6 players)

8
Previous Implementations

• There are many digital versions of Abalone
  available
• The Abalone company offers a free online
  multiplayer computer version on their website
• There are also many freeware and shareware single
  player and multiplayer versions
• But no digital interface between a board and a
  computer could be found

9
Quick fact about the game

• Typical game has only black and white marbles,
  but rule variations allow for up to 6 different
  colors (the colors chosen for this project were
  white and red- to be explained later)
• Using image processing in MATLAB, the marbles and
  open spaces can be detected

10
Different Image Acquisition Options

• We considered many different types of sensors and
  other data gathering devices such as pressure
  and light reflecting sensors, camera sensors, IR
  sensors, etc.

11
Photoelectric Sensors

• Ideal for differentiating between colors
• Small and compact enough to fit into the holes of
  the game board
• Come in different complexities
• Unfortunately, not possible to cost and possible
  reliability issues

12
Pressure Sensors

• Use a silicone top combined with a hydraulic
  system to help determine when pressure changes
• Connected by pins on the other side of sensor
• Unfortunately, there would be too much
  construction (hardware) and cost for this method

13
IR and Light Reflecting Sensors

• These sensors are good in determining different
  colors
• However, they are unable to tell exact location
• Again, cost and complexity are issues

14
Digital Cameras and Webcams

• A very common technology
• Could be mounted above the board at a fixed
  position, or at a variable position
• Cheap
• Would rely on image processing in MATLAB

15
Choosing a webcam

• Cheapest device (25!)
• Pictures can be loaded into MATLAB and analyzed
• Resolution is sufficient to determine location
  and the color of a marble

16
Layout and set up of the board

• For ideal color and location detection, the color
  underneath the holes was made black
• The marble colors being used in a two-player game
  are white and red

17
MATLAB CODE

• Approaches
• Operator independent, automatic detection of
  marbles and holes based on edge detection,
  histograms, thresholding, conversion from color
  to black and white, separation of images into
  red, green, and blue components

18
Examples of MATLAB Experimentation

• Initial image in MATLAB (color)
• Conversion of color to Black and White
• Edge detection (from Black and White scale)
• Not shown histograms, and separation of color
  components

19
Problems Encountered

• Edge Detection
• Works well in detecting different colors (i.e.
  white on black) but does a poor job detecting
  different levels of black.
• Similarity of colors
• Need to enhance the images so that the
  differences in color are more distinct.
• All Image Processing Tools
• Noise problems are very difficult to solve! No
  matter what tools used, it will always be very
  difficult to isolate individual parts of an image!

20
Image Processing Solution

• User initialized board
• Once endpoints are known, all other points can be
  found through midpoint equations
• Processing based on the analysis of the RGB
  values at each point

21
Detection

• After each point is found, image RGB components
  are thresholded to determine marble color and
  space location
• On a 0-255 scale, White has high values for R, G,
  and B, Red has high values of R but not of G and
  B, and Black has low values for R, G, and B

22
Analysis of Moves

• Analysis of moves is based on a logic system
• There are four kinds of moves (push, slide,
  summito, ejection) and multiple directions to
  execute each move
• Ex. A 2 marble slide could be executed in 6
  different directions
• Logic System should check for all possibilities,
  and check for all possible errors

23
Logic System

• Given Boards old and new position (lbig
  11212 dimensional array) created from image
  processing
• Red, White marbles on old and new boards
  (NR1,NR2, NW1, NW2 from lbig)
• of changes and change coordinates (change,
  ccoor(2X array), from lbig)
• How do we determine legal moves?
• Solution 1000 lines of If, ElseIf statements!

24

• lbig(,,1)
• W W W W W
• W W W W W W
• W W W
• R R R
• R R R R R R
• R R R R R
• lbig(,,2)
• W W W W W
• W W W W W W

25

• Now check for ejection
• Check for lt--- and ---gt direction ejection
• elseif ccoor(1,1) ccoor(1,2),
• Check for 2 - 1 ejection
• if (abs(ccoor(2,1) - ccoor(2,2)) 2
  lbig(ccoor(1,1), ccoor(2,1), move) '') ...
• (lbig(ccoor(1,2), ccoor(2,2), move
  1) '' abs(ccoor(2,1) - ccoor(2,2)) 4),
• if lbig(ccoor(1,1), ccoor(2,1), move)
  lbig(ccoor(1,1), ccoor(2,1) 2, move) ...
• (lbig(ccoor(1,1),ccoor(2,1),
  move) 'W' lbig(ccoor(1,2),ccoor(2,2), move)
  'R' ...
• lbig(ccoor(1,1),ccoor(2,1), move)
  'R' lbig(ccoor(1,2),ccoor(2,2), move) 'W')
  ...
• ...
• lbig(ccoor(1,1), ccoor(2,1), move
  1) lbig(ccoor(1,1), ccoor(2,1) 2, move 1)
  ...
• (lbig(ccoor(1,1),ccoor(2,1), move
  1) 'W' lbig(ccoor(1,1),ccoor(2,1), move)
  'R' ...
• lbig(ccoor(1,1),ccoor(2,1), move
  1) 'R' lbig(ccoor(1,1),ccoor(2,1), move)
  'W'),
• disp('A legal 2 - 1 ejection has
  occurred.')
• delete (files1(1).name)
• if NW1 - NW2 1,
• scoreR scoreR 1
• disp('Red has scored.')
• else

26
Results

• Program works! Can referee a complete game
• Continuous image creation through Creative Webcam
  Software and camera mounted to stand
• Program initializes board, keeps track of whose
  turn it is, keeps track of score, determines the
  winner
• If an error is made, program displays last
  correct board orientation and waits for user to
  correct
• Compatible with multiple staring positions
• However, did not have time to make program
  compatible for multiple players/colors

27
Future Work

• Could be adapted for 3-6 players and 3-6 colors
• Further bug testing
• Could implement an artificial intelligence
  program to compete against human opponents (A
  possible new senior project?)

28
Bibliography

• Many meetings with Professor Rudko
• MATLAB, Image Processing Toolbox- users guide.
  The MathWorks Inc, 1997
• Computer Vision Linda Shapiro and George
  Stockman. Prentice-Hall Inc, 2001
• http//www.sensortechnics.com/index.php?fid300fp
  arYToxOntzOjQ6InBjaWQiO3M6MjoiNzIiO303DisSSL0
  aps0blub15a908b8f3eb6ba5eb42abd268d3faa5

29
Any Questions?

• Special thanks to Professor Rudko for his help!