Diploma/Master's Thesis: Vision-Based Analysis of Pool Billiard Games

05.08.2009, Diplomarbeiten, Bachelor- und Masterarbeiten

Project overview: Two cameras are mounted above a pool table to obtain two continuous streams of images from a top-down perspective. The cameras show both the pool table and its immediate surroundings, where the pool players will be situated. The goal of this project will be to design and implement a working system which is able to analyse the current state of a pool game. This includes the number and position of the billiard balls, recognizing players' turns as well as game-ending conditions.

Task description: The student is expected to develop a working system fulfilling the above criteria using hardware and software provided by the chair and formulate a thesis elaborating and evaluating this system. Implementation of the project should be done in C++ to allow for a possible seamless integration with other projects. The HALCON library will be used to provide specialized Computer Vision related methods and a programming interface. Presumably, the more detailed list of tasks expected from the student will be: a) create an abstract representation of the billiard game, which contains information about the state of the board (i.e. position of the balls) and game (e.g. player turns, start/ending conditions). b) synchronize the images captured by the two cameras and integrate them into a seamless view of the pool board. c) use the synchronized image to discover and track the position of billiard balls and integrate this information into the abstract board model. d) discover incidents relevant to the state of the game (e.g. turn changes, balls being pocketed, fouls) and integrate this information into the abstract game model. e) discover game starting and ending conditions (e.g. starting layout, eightball being pocketed) and integrate this information into the game model. The practical work for the thesis will be conducted at the CoTeSys Central Robotics Laboratory (CCRL) in Barer Str.21, downtown Munich. We roughly estimate the practical part of the thesis to take four months of full-time work, leaving two months for elaboration of the thesis.
Prerequisites: General knowledge of Computer Vision algorithms and software development is required. Fair knowledge of the C++ programming language is required for the implementation. The student should be aware that the practical work has to be conducted on site in downtown Munich, and should be willing to cooperate closely with the project advisor. An appropriate workplace will be provided at the CCRL for the duration of the thesis.

Kontakt: eggers@in.tum.de