Tracking Players in Sports Games


Introduction and overview

Application of computer vision technology to sports domain presents a useful and challenging area of research for several reasons. Most sports involve complex human motion and therefore automated capturing, analyzing and quantifying the ability of the athlete can offer significant help to the sports expert.

We concentrate our activities on tracking people in sports based on computer vision. For many years the analysis of a sport event has been based on "observation sheets'' filled-in during the match. In the 1980's, modern techniques of motion analysis were developed with the help of video recordings. Motion acquisition and analysis were performed manually, which was a time consuming and tedious task. In the past, progress in introducing the computer vision technology to the team sports domain was slow, mainly due to inadequate video and computational facilities, as just a single match may require processing of tens of thousands of complex images.

Large amounts of data and high computational load are by no means the only burden. The players strive to move rapidly, change direction unpredictably and collide with one another. They violate the smooth motion assumption, on which many tracking algorithms are based. Players appear in the images as highly non-rigid forms, especially due to the movements of their extremities. In addition, cameras used to record the sport events have to cover large area, either by following the players of interest, or by using wide-angle lenses, which results in substantial image distortion and low resolution. Many of the proposed approaches solved the motion acquisition and analysis problem only partially and were therefore unable to provide an adequate solution to the sports experts, i.e. tracking every player and/or the ball in the whole field, and in every instance of time.

This research project is a result of cooperation with Faculty of Sports at University of Ljubljana and is currently divided into the two parts, as different approaches are needed in dealing with different types of sport.

Part 1: Team sports (handball, basketball)

Part 2: Individual sports (squash)

Handball video demonstration (input data, processing and results - 21 MB, 60 sec, MPEG)  


Recent Work:

Continuing the work on human motion analysis, we have developed between years 2003 and 2008 a new version of our computer-vision-based system for sports analysis. The system is composed of four major parts:
(i) Camera calibration module, which is highly intuitive and easy to use tool for establishing a mapping between video data and ground plane of the court;
(ii) Tracking module which allows semi-automatic tracking of a particular player or a group of players during a sports match;
(iii) Presentation module is used for visualization of the obtained data;
(iv) Video export module can be used to generate annotated virtual-camera videos of selected players;
(v) Activity recognition module, which is used for data-mining and tactical analysis of sports matches. More on activity recognition can be found here.

The tracking and video exporting engines are written in C++ with native interfaces to Java user-friendly interfaces, which makes the application fairly platform-independent (tested under Linux as well as under Windows). This application has been extensively used by sports experts from the Faculty of Sport University of Ljubljana, and by sport experts from Ruhr-Universität Bochum, for motion analysis of players during basketball, european handball, squash and tennis. The sports-oriented research so-far has been primarily focused on answering some fundamental questions about a particular sports discipline, such as the expected distances travelled per match, in which part of the court does a player spend most of time, how do player's velocity and acceleration vary during a match, what are the major differences in the dynamics of losers and winners, etc.

A video describing the capabilities of the tracking system was accepted to 10th European Conference of Computer Vision (ECCV08) and can be found below;

Online presentations:

Tracking of players in the case of handball (methods)

Error analysis, accuracy evaluation and testing.


Selected publications:

M. Kristan, J. Perš, M. Perše, S. Kovačič: "Closed-world tracking of multiple interacting targets for indoor-sports applications", Computer Vision and Image Understanding, 2008 in press.
M. Perše, M. Kristan, S. Kovačič, G. Vučkovič and J. Perš: "A Trajectory-Based Analysis of Coordinated Team Activity in a Basketball Game", Computer Vision and Image Understanding, 2008 in press.
J. Perš, M. Bon, S. Kovačič, M. Šibila, B. Dežman: "Observation and Analysis of Large-scale Human Motion", Human Movement Science, 21(2), 295-311, July 2002.


J. Perš, S. Kovačič, "Tracking People in Sport : Making Use of Partially Controlled Environment" In: SKARBEK, Wĺadysĺaw (Ed.). Computer analysis of images and patterns : 9th international conference, CAIP 2001, Warsaw, Poland, September 5-7, 2001: proceedings, (Lecture notes in computer science, 2124). Springer, 2001, pp. 374-382. (note: the PDF file contains the expanded version of the paper!).
J. Perš, G. Vučkovič, S. Kovačič, B. Dežman, "A Low-Cost Real-Time Tracker of Live Sport Events." In: LONČARIĆ, Sven, BABIĆ, Hrvoje (Eds.). ISPA 2001 : proceedings of the 2nd international symposium on image and signal processing and analysis in conjunction with 23nd int'l conference on information technology interfaces, Pula, Croatia, June 19-21, 2001. University Computing Center, University of Zagreb, 2001, pp. 362-365.
J. Perš, M. Bon, S. Kovačič, "Errors and Mistakes in Automated Player Tracking." Proceedings of the Sixth Computer Vision Winter Workshop, pp. 25-36, Bled, Slovenija, February 7-9 2001.
J. Perš, S. Kovačič, "A System for Tracking Players in Sports Games by Computer Vision." Electrotechnical Review - Journal for Electrical Engineering and Computer Science, 67(5):281-288, 2000.
J. Perš, S. Kovačič, "Analiza in prikaz rezultatov sledenja igralcev v ekipnih športih", Zbornik devete Elektrotehniške in računalniške konference ERK 2000, zvezek B, 21.-23. september 2000, Portorož, Slovenija, uredil Baldomir Zajc - Ljubljana: IEEE Region 8, Slovenska sekcija IEEE, str. 261-264, 2000
J. Perš, S. Kovačič, "Computer Vision System for Tracking Players in Sports Games", Proceedings of the First Int'l Workshop on Image and Signal Processing and Analysis IWISPA 2000, pp. 81-86, June 14.-15. 2000, Pula, Croatia.
J. Perš, S. Kovačič, "Sledenje igralcev v športnih igrah z računalniškim vidom" ("Computer Vision System for Tracking Players in Sports Games"), Eighth Electrotechnical and Computer Science Conference ERK'99, Portorož, Slovenija, 23-25. September 1999.