George Campbell David HawesCarleton Jillson Joseph KalinowskiKeith Pray CUE The Pool Critic

Project Description The goal of our project was to build a system which would model a live pool game. It would then analyze the model to critique each shot.

Project Design Table State Look Image Video InGet Status Status Video Source Movie Vision Analyzer User Interface

Project Design User Interface Analysis Transformation Vision

Project Design User Interface Analysis Transformation Vision

User Interface Designed using MFC Multi-Document Application Easy to use for the novice and expert Intended to have minimal back-end communication

User Interface

Project Design User Interface Analysis Transformation Vision

Analysis The analysis is responsible for building and interpreting the model of the live pool game. Model Parts –Table State –Events –Shot

Analysis Vision User Interface Table States Events Called Shot Result

Project Design User Interface Analysis Transformation Vision

Transformation Using vanishing points, a video image of the table is divided into regions which are mapped to relative regions of a table model. Perspective issues: –Room constraints –Program more robust Perspective issues: –Room constraints –Program more robust

Transformation

Project Design User Interface Analysis Transformation Vision

Acquiring Video Vision can accept any of the following formats –Live Video –AVI Files –Directory of static images

Vision Issues Vision uses the vast majority of the process time –In order to have acceptable detail, large numbers of pixels must be processed –Very efficient algorithms needed to produce accurate results without slowing down CUE to a crawl

Noise needs to be compensated for: –Reflectivity of balls –Shadows –Camera Noise Vision Issues

How Does Vision Work? Calibration Processing each Image

Calibration Using one image from the image source Find the edges of the image Find the key lines (bumpers) of the table Map the four corners of the table to model space using lookup table module

Calibration Image

Processing Images Find the edges of the image Find the circles on the table Identify balls based on color matching Set the table state with ball locations Return the table state for analysis

Sample Edge Image Sample Processed Image Processed Images

Demo

Future Projects Incorporate AI Aspects –Learns physics/rules of pool by watching games in play –Uses strategy to critique and recommend shots Robotic Implementation –Learns by trial and error –Carries out recommended shots in a real game

Special Thanks To... Professor Michael Gennert Spencer Billiards John Chaillet Lisa Cocozzella Worcester Telegram & Gazette Thank You

George Campbell David HawesCarleton Jillson Joseph KalinowskiKeith Pray CUE The Pool Critic