Tracking people in sports using video analysis
(Tracking von Personen im Sport unter Einsatz von Videoanalyse)
Analysis of movement in sport events can be of high importance to professional athletes, and video recordings of sport events are widely available. In this project, we wish to utilize the video information for the movement analysis, which leads us to the following main question:
How can computer vision be used to analyze movement in sport events?
How can we automatically find the players in a video clip? Can we find the path of one player's movement through a video clip?
Can this method be extended to finding paths of several player's movement in the same video clip?
How can we distinguish players from each other if they overlap in the image?
Can we find the position of the players on the field in real world coordinates?
What kinds of sports are optimal for using the methods developed?
How can the methods we develop be generalized to accommodate more
sports?
We have chosen to use Matlab as the tool for analysis and calculations. Matlab is very well suited for doing concept modeling and prototyping of mathematical applications, which this project is all about, and it enables us to go directly into the interesting video analysis parts instead of spending time setting up code to handle pixels, images, and video les. We do not want to create an on -the- shelf product ready for distribution, nor will we spend resources and time on creating a fancy GUI. Instead we will focus on creating Matlab functions and Matlab demo scripts, that can be used as proof-of-concept and could be further extended or later used as the engine in a GUI based application.
One downside of using Matlab is the poor performance compared to a low-level programming language as C. If we were to create a system that could do tracking in realtime, we were forced to use a low-level programming language, and maybe even to create some dedicated hardware for the task. This is way out of the scope of this project, so therefore we chose to ignore the poor performance and go with Matlab.
© Copyright 2008 Veröffentlicht von Technical University of Denmark. Alle Rechte vorbehalten.
| Schlagworte: | |
|---|---|
| Notationen: | Naturwissenschaften und Technik Spielsportarten |
| Sprache: | Englisch |
| Veröffentlicht: |
Kongens Lyngby
Technical University of Denmark
2008
|
| Online-Zugang: | http://orbit.dtu.dk/getResource?recordId=220921&objectId=1&versionId=1 |
| Seiten: | 86 |
| Dokumentenarten: | Diplomarbeit |
| Level: | hoch |