Tennis Player and Ball
-
Upload
lau-yeowhong -
Category
Documents
-
view
220 -
download
0
Transcript of Tennis Player and Ball
-
8/15/2019 Tennis Player and Ball
1/22
Tennis Player and Ball
TrackingYeshwanth Malekar
-
8/15/2019 Tennis Player and Ball
2/22
CONTENT
MotivationChallenges
Previous Work
Algorithm
DetectionSolution
Problems
Results Future Work
References
-
8/15/2019 Tennis Player and Ball
3/22
Motivation
Tracking Tennis players and ball in a Tennis ma
Useful for automatic tennis video annotation.
Player and ball trajectories contain high levelinformation which can be used for training pudesigning strategies.
-
8/15/2019 Tennis Player and Ball
4/22
Project Goal
-
8/15/2019 Tennis Player and Ball
5/22
Challenges
Video taken is not from a steady camera.
The speed and size of the ball pose real probl
Most of the times, ball undergoes either occlublurred in the background.
Chances of false detection are high.
-
8/15/2019 Tennis Player and Ball
6/22
Previous Work
Most of the previous work is done on a video ta steady camera.
They also used Kalman filter and Particle filtersthe location of foreground objects.
These filters need a steady camera to work onmodel of the movement of players and balls.
-
8/15/2019 Tennis Player and Ball
7/22
Algorithm
1. Extract Background image.
2. Frame difference method.
3. Morphological Operations.
4. Blob analysis.
5. Detect players and ball.
-
8/15/2019 Tennis Player and Ball
8/22
Background extraction using median function.
Bad background examples:
Too few Frames: Over all frames:
-
8/15/2019 Tennis Player and Ball
9/22
Foreground Separation Morphological Operation
Player and Ball detection
-
8/15/2019 Tennis Player and Ball
10/22
Conditions
Added extra boundary conditions for the ball
Finally centroid conditions to keep track of theand ball in the following frames.
-
8/15/2019 Tennis Player and Ball
11/22
What if camera moves?
The whole court is detected in thedifference image instead of justForeground.
-
8/15/2019 Tennis Player and Ball
12/22
Simple solution
Skipped through the frames with more than required blobs.
Continue detecting the players when the video is stabilized.
Saved the centroid of the players detected in all the frames.
Adjustments:
Added the difference of the centroid positions of last two imaprevious image.
We have new approximated candidate positions for the blan
-
8/15/2019 Tennis Player and Ball
13/22
Advantages of this solution
It is very simple technique to implement.
Players detection is excellent.
Even if the ball is occluded or blurred, it is dete
-
8/15/2019 Tennis Player and Ball
14/22
Advantages (Cont.)
-
8/15/2019 Tennis Player and Ball
15/22
Problems encountered
The ball moves at high speeds and changes itrapidly.
If detection in a frame is missed, keeping trackball in the next frame with this approximation not possible.
False detections influence further false detect
-
8/15/2019 Tennis Player and Ball
16/22
Problems (Cont.)
-
8/15/2019 Tennis Player and Ball
17/22
Results
-
8/15/2019 Tennis Player and Ball
18/22
Results (Cont.)
-
8/15/2019 Tennis Player and Ball
19/22
Result (Doubles)
-
8/15/2019 Tennis Player and Ball
20/22
Future Work
Better ball detection techniques for a moving
Would also like to work on different ball gamesoccer.
-
8/15/2019 Tennis Player and Ball
21/22
References
Hira Fatima et al, “Object Recognition, Tracking and Trajectoin Real-Time Video Sequence”, International Journal of InformElectronics Engineering vol.3, no.6, pp. 639-642, 2013.
Amor Salorpour et al, “Vehicle Tracking Using Kalman Filter anSignal & Image Processing, an International Journal (SIPIJ), VJune 2011.
F. Yan and W. Christmas and J. Kittler, “Tennis Ball Tracking AlgAutomatic Annotation of Tennis Match”, BMVC, pp 619-628, 2
-
8/15/2019 Tennis Player and Ball
22/22
Questions