Scientists Claim Slam-Dunk with New Sports Analysis Software

Category: Technology / Mar 22, 2012 2:27PM EDT

Scientists in Switzerland have developed software that allows computers to track the real-time movements of basketball players using multiple cameras.

The software is sure to interest those in the professional game, as it allows coaches to improve their monitoring of individuals, both on their own side and the opposition's.

It's been developed by computer scientists at the Ecole Polytechnique Federale de Lausanne (EPFL), led by Professor Pascal Fua.

Working in the EPFL's Computer Vision Laboratory (CVLab), the team use multiple, precisely arranged, video feeds that are constantly analyzed by three individual algorithms.

The system is made up of eight standard cameras: two on each side of the field or court; two that film from above; and a further two that zoom.

The first algorithm divides the playing area into grids of just 25 centimetres squared, simultaneously removing the background and deducing the probability of the presence of a player in each of the small squares.

The second algorithm handles player tracking, estimating the location and movement of players based on the work of the first algorithm.

Since a computer can confuse or lose a player and his or her data during movement and contact, the final algorithm system monitors, records and follows the players' appearance, including the colour of their shirt, their number and other visual cues.

PhD student Horesh Ben Shitrit is improving the original tracker prototype and his final adjustments should create a system capable of tracking players reliably in real time.

Our system is able to track multiple people from multiple cameras. We can estimate the position of each player at each time on the floor and from that we can have the trajectory of each player during the match. We use the trajectory in order to do analysis of the behaviour of the players and the behaviour of the teams and we can improve the performance of the teams and the players, said Shitrit.

Ben Shitrit believes basketball coaches would benefit enormously from the technology. The system might not be finished in time for this Summer's Olympics but, if it were ready says Ben Shitrit, coaches at the Games would benefit from the technology.

You want to do, want to know, that you did precisely the right thing that you were aiming to and you want to understand your team-mates during the match and if you could start analysing your behaviour before the Games, the Olympic Games, then while the Olympic Games are running you could understand what the other teams are doing, he said.

 The system could potentially be used to track various team sports, including soccer and American football.

Professor Fua believes the system could also be useful for surveillance, crowd control -- and even as a tool to help architects.

You could watch how a building's being used. Typically, when an architect builds a building he has a particular pattern of usage in mind, but who knows?. Maybe it's being used differently and a kind of a system like this could be used to  do this and it could be used to do this without intruding on the privacy of people, because we don't have to record their identity in the end. All what you have is trajectories, it doesn't say who they are, said Fua.

The EPFL team says its technology could also be used to create computer sports commentary.

Fua believes sports coaches will be particularly interested in the potential of in-depth study of opponents.

You could analyse the tactics of your team and, even more interestingly, the opposing team and maybe try to watch them play in various games and try to understand what their strategy is, what their play book is and that sort of things, and that's things that I think coaches for all team sports will be interested in, he said.