Automatic)Score)Keeper)for)Cornhole–Senior)Design

Transcription

Automatic)Score)Keeper)for)Cornhole–Senior)Design
Automatic)Score)Keeper)for)Cornhole–Senior)Design
Brock)Germinara,)Tologon)Eshimkanov,)Nate)Derbinsky,)John)Russo,)Durga Suresh
Wentworth)Institute)of)Technology
College)of)Engineering)&)Technology,)Department))of)Computer))Science)&)Networking
Introduction
Problem5&5Solution
Development
Results
Cornhole
Problem
Program flow design
Conclusions
• Game consists of two boards, each with a “corn hole”
• Each team has four bean bags with designated colors
• Typically two players per team= one from each team at
either board
• Each player throws bean bags onto the opposite board
• Need of knowledgeable person(s) to keep score
updated for each team
• Point above takes time and human resources
• The score can be calculated based on an image from
the dataset
• The score can be calculated using a custom image
• The score can be calculated given an open video
stream
•
•
•
•
The primary usage of score keeper is computer vision
The score keeper relies on image preZprocessing
Illumination plays an important role in overall detection
The score keeper is effective (90 and 85 percent
accuracies) without need to rely on tools such as
machine learning
Bibliography
Data>flowLdiagram
Input
Cornhole gameLinLaction
Rules
•
•
•
•
A bag on the board provides one point
A bag going into the corn hole awards three points
A bag off of the board gets zero points
The round ends after each team throws four bean bags
from one side to the opposite board
• After each round, the difference in scores is added to
the team with the best score for that round
• Game ends when either team reaches a total score of
21 points and is ahead by 2 or more points
Solution
• Automatic score keeper
• Use of video or static images to observe the board
• Display scores for each team at any moment
Accuracy metrics
• Given the dataset, metrics measure the accuracy of
the automatic score keeper
• Color metrics explained color detection accuracy
• Location metrics provided insights into bean bag
on/off/in location of each bag compared to the board
• The score keeper achieved ~90% color accuracy
• The score keeper produced ~85% location accuracy
ColorLinaccuracy
GameLscoringLrulesLrepresentedLasLtheLgraph
Output
LocationLerror
1. "How to Use Background Subtraction Methods." How
to Use Background Subtraction Methods — OpenCV
3.0.0>dev Documentation. N.p., n.d. Web. July 2016.
2. "OpenCV: Hough Circle Transform." OpenCV: Hough
Circle Transform. N.p., n.d. Web. July 2016.
3. "OpenCV: Image Thresholding." OpenCV: Image
Thresholding. N.p., n.d. Web. June 2016.
4. "Smoothing Images." Smoothing Images — OpenCV
2.4.13.0 Documentation. N.p., n.d. Web. June 2016.
5. "Finding Contours in Your Image." Finding Contours in
Your Image — OpenCV 2.4.13.0 Documentation.
N.p., n.d. Web. June 2016.
Acknowledgment
First and foremost, we express our gratitude to Dr.
Nate Derbinsky, a professor in the Department of
Computer Science and Networking. He is in charge of a
research project where a robot, “CHUCK,” is being
developed to play full Cornhole games. He gave us an
opportunity to work on a project that will help CHUCK
observe a platform and build its game strategy.
Professor Derbinsky mentored the development of the
automatic score keeper and provided valuable
feedback.
We also would like to thank Tyler Frasca, for his
developments in CornholeZrelated software, specifically
the dataset creation. Tyler’s initial work was a
foundation upon which the automatic score keeper was
built. The dataset helped to validate our software and to
continuously improve it.
www.postersession.com