Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 43769 (19)
- University: Sharif University of Technology
- Department: Computer Engineering
- Advisor(s): Kasaei, Shohreh
- Abstract:
- By progress of computer vision, many systems are produced for soccer analysis. These systems have been used in many applications such as 2D and 3D reconstruction in arbitrary view. In this circumstance, we need to detect and track players, referee and ball in whole matches. In a single view system there are problems such as occlusion. By having more views of game, when two objects are occluded in one view, we can detect and track them in the others. Objective of this project is multi-view detection and multi-view tracking players and ball in soccer. In recorded videos from multiple views, we track path of players and the ball.
Proposed player tracking system has two main parts. First part which called single view tracker, detects players’ position in each view. Then using 3D geometry calculations players’ position in top view is achieved. Second part which called central tracker, tracks players in top view by Kalman filter. Ball tracking system finds possible balls in each frame of single view by means of its features. Then a number of 3D possible balls are obtained. Finally, using of continuous ball path and combination of information from each view, 3D ball path is determined.
Performance of proposed system is examined on a recorded dataset from 9 fixed cameras in Azadi arena. Position of the ball found correctly over than 90% of frames. Also, tracking of players has been done correctly in 96% of frames - Keywords:
- Location ; Tracking ; Football ; Multiview Tracking
Digital Object List
-
محتواي پايان نامه
- view
Bookmark