TU München Department of Computer Science Intelligent Autonomous Systems
Home About Technology Videos Publications Team IAS@TUM

 
Input Video #1 Input Video #2 Input Video #3 Suat Output Video #1 Output Video #2 Output Video #3 Output Video #4 Output Video #5 Output Video #6

ASPOGAMO Technology

Watching Football through the Eyes of Experts

The ASPOGAMO system includes a tracking system that determines player trajectories in camera views based on TV broadcasts. To do so, ASPOGAMO solves a complex probabilistic estimation problem that consists of three subproblems that interact in subtle ways:

  • estimation of camera direction and zoom factor,
  • detection of players and team affiliations and tracking of multiple players, and
  • disambiguation of tracked players after occlusions

The Aspogamo game perception component consists of four components:

  • Vision,
  • State Estimation,
  • Track Editor, and
  • Motion Model Generator

The Vision is responsible for image processing tasks and the State Estimation is used to estimate camera parameters and player positions as well as forming player tracks.

ASPOGAMO processes a video-stream by performing the following steps for each image: First, the Camera Parameter Estimation predicts camera parameters for the current frame using optical flow from the Correspondence Finder. Second, the Player Detection determines player blobs taking into account the hypotheses generated by the Tracker. Third, knowing the camera parameters, player positions and their uncertainties are calculated in the Player Position Estimation from the meassurements made by the Player Detection. They are forwarded to the Tracker, where individual positions are combined and associated over time, and hypotheses for the next time step are generated.

 
 

Computer Science IX | Technische Universität München