Table Tennis is a popular game. This project suggests a way to analyze the match using Kinect camera. During the project an algorithm was developed which get an input video of the table tennis match, tracking the ball and the player, identifies the bounds of the table and classify every stroke if it is a Backhand ot Forehand stroke. In addition the algorithm recognizes the time of the ball impact with the table/racket.
Abstract
Table Tennis is a popular game. This project suggests a way to analyze the match using Kinect camera. During the project an algorithm was developed which get an input video of the table tennis match, tracking the ball and the player, identifies the bounds of the table and classify every stroke if it is a Backhand ot Forehand stroke. In addition the algorithm recognizes the time of the ball impact with the table/racket.
Algorithm
Tracking users with Kinect skeletal tracking
Classify backhand and forehand
Conclusions
- Tracking users with Kinect skeletal tracking
-accurate
- Identify the table
-accurate
- Tracking the ball
-The Kinect depth sensor doesn’t identify the ball in a reliable way
-On average we found the correct ball position in 90% of the cases
- Identify ball events –
-We are able to determine the frame in which the ball hits the table and the racket
Ideas for future research
- Use two kinect cameras, when each camera tracks one player.
- Use the ball events identification to improve the ball tracking
- Use high speed camera instead of kinect
- Use advanced depth camera instead of kinect