Pool School: Color-based Object Tracking and Shot Prediction as a Learning Tool in Billiards

Abstract

Table sports are one of the oldest and most cherished class of games, and as long as they have been around, they have inspired competition. Many methods for improving one's shot accuracy exist in today's market, but none specific to each unique placement of the balls on the table. The objective of Pool School was to create a system which identifies the locations of all balls on the board, calculates the optimal angle of attack, and projects it onto the board so that one may align their cue and get the ball in the pocket. It is first and foremost a learning mechanism.

Pool School is a capstone Senior thesis, combining foundational mechanical engineering curriculum and methods. This paper describes the design, development, and results of Pool School over the 2017-2018 academic year. The design of Pool School may be split into four major sub-assemblies: the table and mounting system, the camera, the predictive code, and the projector. Due to the low error margins allowable in shot prediction, the camera and projector systems were calibrated and evaluated to minimize distortions from tilt, lens design, and aspect ratio. The mounting system was designed and manufactured to accommodate the table size and throw ranges of the projector and camera as well as their weight. The system in entirety was then evaluated for percentage angular error of predicted paths and these results were compared to the permissible ranges calculated in the problem statement.

For shots within the center $54 degrees of the pocket measured in testing and the core view of the camera, the system functioned within a 1.5% error margin. Despite its limitations, the final system achieved the baseline objective set out to be accomplished in this thesis.