Original URL: http://www.theregister.co.uk/2010/07/10/goalline_technology_worldcup/

Image recognition – defense against a Lampard replay?

Geo-location missing an open goal

By Gavin Clarke

Posted in Science, 10th July 2010 07:02 GMT

The inventor of image recognition technology used in sports across the world has written to international football's governing body to ask that they reconsider using Hawk-Eye, a tool capable of spotting when a ball has crossed the goal line – and avoiding a repeat of the England versus Germany World-Cup debacle.

Speaking ahead of this weekend's Netherlands versus Spain final, Paul Hawkins - managing director of Hawk-Eye Innovations - has yet to get a reply from FIFA.

His letter comes ahead of a meeting by the International Football Association Board (IFAB) at the end of this month, where use of goal-line technology like Hawk-Eye will be discussed. IFAB sets football's rules and says what technology is allowed in matches.

Football's lawmakers rejected use of goal-line technology before the World Cup – in spite of IFAB approving use of a modified version of Hawk-Eye in a pilot three years ago at Reading FC.

That was before England midfielder Frank Lampard's goal was disallowed against Germany during a World Cup round-of-sixteen match in South Africa, in a match Germany won 4-1.

The linesman might not have seen the goal, but it was clear to the rest of the viewing world what had happened.

When FIFA announced its decision not to use goal-line technology, general secretary Jerome Valck said he hoped it would not come back to haunt them.

The haunting didn't stop at England versus Germany. Mexico fans watched horrified as Argentina's Carlos Tevez scored a goal that was clearly offside – clear to anyone but match officials. The goal helped end Mexico's world cup challenge.

Hawk-Eye's inventor reckons you can solve offside too. Hawk-Eye could track players in the same way it tracks balls by combining the data to determine a position. "That's technologically achievable – we could do it," he said.

The problem is that the work takes funding, and nobody wants to press on given IFAB and FIFA's apparent objection to use of on-pitch technology.

The reason for this is unclear, especially given that IFAB OK'ed the Reading pilot of Hawk-Eye in 2007 and 2008. The pilot was paid for by the English Premier League, which has come out in favor of goal-line technology.

A common objection is it might spoil the game by creating stoppages in play. Others say the technology might not be accurate.

Hawkins has no time for such complaints, claiming Hawk-Eye can get results from the goalmouth into a ref's ear in half a second. "Anyone who makes that argument doesn't know what they are talking about," Hawk-Eye's inventor told us Friday ahead of the final.

Hawk-Eye has proved its accuracy in tennis, cricket and snooker, and Hawkins modified the system for the Reading pilot once he got the clearance from IFAB.

The football systems uses six cameras around the goalmouth to record the action, with each camera hooked up to a server to capture and process data from the frames. Data is fed to a central server and knitted together to triangulate that position of the ball to a distance of within 5mm.

The image scanning and processing software is built in C++ and at Reading ran on Windows and Dell servers, with algorithms tweaked so the software could account for more people in a tighter spot – the penalty box and goal mouth – and differentiate, for instance, between the curve of a ball and that of a number "2" on a player's shirt.

The Reading test passed with full marks, catching thousands of goal-line incidents.

"Ultimately the software processes the images to always find the ball even if part of ball is obscured,' Hawkins said.

If IFAB and FIFA are against HawkEye's image recognition, they have few other options – even in the age of location-based services.

Anyone for GPS?

It would seem that if you can track an iPhone or a stolen laptop using GPS or post your exact whereabouts in a massive city like New York on your phone using Glympse, you could track the position of a ball and 22 players in the relatively small area of a football pitch. Load them up with GPS tracking, and you're off.

Unfortunately, while GPS services can give your general position in a Times Square, they aren't accurate enough for the beautiful game.

GPS needs a line-of-sight to the satellite overhead. Otherwise, you get inaccurate and delayed results as the signal is obscured by buildings or gets bounced around. GPS is accurate to within 50 to 100 feet.

Ted Morgan, chief executive of location-based service provider Skyhook, which powers the iPhone and devices from Dell, Samsung and Motorola, said football would be the worst of all scenarios for GPS. The desert just outside Baghdad is ideal because the lack of structures – and people.

"Human beings are the worst thing to have around because they are full of water - they make the signal bounce like crazy and you have stadium full of metal that acts like a satellite dish. Open fields are awesome," Morgan said.

GPS can be boosted using a couple of techniques. You can triangulate a device's location using information from nearby cell-phone towers – called Assisted GPS. Also, you can use Wi-Fi with a device measuring the strength of signal from Wi-Fi transmitters and triangulating based on the identity of the Wi-Fi transmitter that the receiver has passed near. Skyhook takes advantage of Wi-Fi, GPS satellites, and cell towers plus its own positioning algorithms – and still it's looking at between 30 and 60 feet.

Geoff Glave, product manager with Absolute Software that tracks stolen PCs using GPS and Wi-Fi, puts Wi-Fi's accuracy at between 100 and 200 feet. "If you are holding a device and move 14 inches you are not going to see it - not until you walk a couple of feet will you see a change," Glave said.

Wi-Fi has an added complication. For more accuracy you must map the venue to try and pinpoint an object – a timely and potentially costly exercise that can militate against its use. But again, you're talking about a range of nine to sixteen feet in accuracy, says Skyhook's CEO.

RFID tags could be another potential answer. However, tags are only registered when they are near sensors, so you'd need to cover the ball and players with tags and the pitch with sensors.

Morgan is horribly realistic about the chances for GPS for the foreseeable future. Researchers are chipping down the yards, but the sweet spot for geo-location remains consumer devices where you only need an approximate location – such as the position of a hotel or a restaurant on a street.

"I think technology is overkill. They should try the tennis stuff - that works pretty well," Morgan said of football. He didn't name it, but the chief of a GPS and WiFi positioning company relied upon by millions of iPhone and Android users to plot their positions was talking about Hawk-Eye. ®