|
Overview
In presented work, we focus on tracking the participants (players) in
individual sports. In contrast to team sports, there are only a few players
visible in the observed scene. Therefore more basic tracking algorithms can be
employed, which results in fast (near real-time) image processing. The system
presented here was developed by using the recordings of a squash matches,
however, it could be adapted to many other sports (for example tennis,
one-on-one basketball and similar). The whole match is videotaped using single
stationary camera. The video is digitized to M-JPEG format using inexpensive
consumer-grade video acquisition hardware. Player tracking and motion analysis
are done offline. The system for tracking players in individual sports been developed to the phase of working prototype, which is routinely used at Faculty of Sports at University of Ljubljana for squash analysis. To enable routine use, an user interface, which is shown below, was developed. The application runs under the Microsoft Windows operating system.
Simple algorithms for player tracking are implemented in this prototype, which enabled us to use highly optimized Intel Image Processing Library for video data manipulation. This results in near real-time processing speed, which reaches 15-20 frames per second on Intel Pentium III class PC. System is supervised by human operator during the tracking process, but needs little intervention if proper precautions were taken before and during the video recording. This way, several hours of sport video can be processed in single workday.
Results, provided by our system can be exported in the form of tab-separated text files, which ensures easy integration with other applications, which can be used for further analysis. Alternatively, results can be presented in illustrative and visually attractive manner, for easier analysis by sport experts. Some examples are shown below.
Contact |