Action Detection in Ball Sports

Actions in a Soccer Exercise — Automatically detected ball contacts, passes, dribblings, paths and shots of three tracked players participating in an attack.

Project leader: Dipl.-Ing. Nicolas Witt
Project members: Dipl.-Ing. Nicolas Witt
Start date: 01.08.2016
End date: 31.12.2018

Abstract

Tracking devices based on GPS are often used to assess an athletes physical performance. New sports tracking technologies might also enable a technical assessment through higher sampling rates, precision and the ability to track both players and the ball. The RedFIR system of the Fraunhofer IIS is a so called RTLS (real-time tracking and locating system) an uses small transmitters for tracking athletes and the ball.

The goal of this project is to detect actions in training exercises and matches using tracking data.
This mainly enables the following applications:

Technical assessment of players in training sessions (e.g. time in ball possession)
Automatic generation of exercise animations
Interactive training through online action detection and feedback
Competing against the best execution (e.g. compete against ghosts)