Cameras and vision algorithm track athletes
A new system makes it possible for spectators to follow multiple players at once on a football field or basketball court, even when they are crouched behind another player.
The system—developed at École Polytechnique Fédérale de Lausanne's (EPFL; Lausanne, Switzerland) Computer Vision Laboratory (CVLab)—comprises eight cameras: two on each side of a field or court, two that film from above, and two that can zoom in on individual players.
Once the image data have been acquired by the cameras, they are processed on a computer using algorithms that detect and mark the location of individual players.
The playing area is first sliced into small 25 × 25-cm squares, after which the background in all the images is simultaneously removed. Next, the software deduces the probability of the presence of a player in each of the squares. The results are then used to establish individual trajectories of the players. Last, the color of the uniform and the unique number on each shirt is used to classify each player.
After the location and identity of the players have been computed, an image is then superimposed on top of them, showing the color of their jerseys and their number so spectators, referees, and coaches can easily follow individuals without mixing them up.
Once the system has been further enhanced to allow television cameras to provide direct input to it, the researchers say that it could be used for international sporting competitions.
But the researchers believe that the system could also be used outside the sports arena, too. They plan to extend its functionality to allow it to track pedestrians and follow the movement of clients in a store for marketing purposes.