Levante Savo, Camelia Popa,

jabgoldfishAI and Robotics

Oct 19, 2013 (3 years and 5 months ago)

76 views

The Computer Vision Game Assistant


C


H


E


C


K


E


R


S

M


A


T


E


S

Levante Savo, Camelia Popa,

Serban Meza, Oscar Medina,

Ross Foley


Build a Machine Vision System that:

1.
Reads the current state of a board game using a
webcam

2.
Feeds the current game state into a game
playing engine

3.
Retrieves a suggestion for the next move


1


Game: Checkers/Draughts


Very popular


Many game playing engines
available on the web


Pieces relatively easy to
segment



Programming Language:
Matlab


Powerful IP functionality


Direst access to webcam
using Simulink


Simple access to game
playing engine .exe file



2

3


Image taken every 2
seconds


Movement Detection


Compares image to
reference image


If large difference,
presume a hand is
obstructing the view and
no game is extracted


If small difference, give
image to the game
extraction module and this
image becomes the
reference



4

Board & Piece Detection & Combination

5


C


H


E


C


K


E


R


S

M


A


T


E


S

6

7


Centroid positions combined with box
information


8x8 matrix created


Each entry represents 1 box on the board


Contains flags indicating whether box is not used,
empty, black/white, man/king

10

Query Generation & Results Visualisation


11


C


H


E


C


K


E


R


S

M


A


T


E


S

12


Found on the web


Application written in C++


Takes 8x8 matrix as input


Outputs 2 co
-
ordinates, starting position and
finishing position


Also a string detailing an error or if a player
has won/lost

13

14

15