Homegrown math scores in tiki-taka tracking

ISI team develops algorithms that promise higher accuracy than the competition

The automated system developed at ISI looks to be as accurate as Lionel Messi’s left foot when it comes to ball tracking. (AP)

Dunlop:

Advanced algorithms developed by a team of scientists at the Indian Statistical Institute (ISI), Calcutta, promise to do for ball tracking in football what tiki-taka has done for FC Barcelona.

Dipti Prasad Mukherjee of the institute’s electronics and communication sciences unit and PhD candidates Saikat Sarkar and Samriddha Sanyal have come out with a high-precision automated system using algorithms that “balance prior constraints continuously against the evidence garnered from sequences of images”.

The system, based on computer vision, aims to track and accurately calculate ball possession by a team during a game. According to Samriddha, it is at least 7.2 per cent better compared to competing approaches to ball tracking.

Samriddha had published a paper titled “On the (soccer) ball” with Mukherjee and Arnab Kundu in 2016.

“The problem of tracking the ball is when there is a sudden change in speed and orientation of the ball,” he said.

In a given video, the prior constraints would be the ball positions in previous frames. The paper proposes a particle filter-based algorithm that tracks the ball when it changes direction suddenly or travels at high speed.

“Our tracking algorithm has shown excellent results even for partial occlusion (blockage), which is a major concern in soccer videos,” Samriddha said.

Tracking a ball when it is being kicked or passed quickly from player to player like in tiki-taka remains a challenge for broadcasters.

Saikat, who is working on calculating ball possession, said: “Till now, the chess-clock method is used to measure ball possession. The other measure for ball possession is to count the number of passes. The ratio of the number of passes by a team divided by the total number of passes in a match closely correlates with ball possession stats.”

FIFA uses data from Deltatre, a sports media company that uses the chess clock method. In the Premier League, ball possession is measured on the basis of data from Opta Sports.

Deltatre has individuals using the three buttons of a chess clock to measure when the ball is with Team A, Team B or not in play. Opta uses a software overlaid on live feeds to track the number of passes.

The team at ISI uses raw broadcast video, measuring ball possession with close to 80 per cent accuracy.

source: http://www.telegraphindia.com / The Telegraph,Calcutta,India / Home> Calcutta / by Anasuya Basu / March 14th, 2018

Leave a Reply

Your email address will not be published. Required fields are marked *