Technology Meets Tennis, or the (Machine Vision) Eyes Have It
Close only counts, we've been told, with horseshoes and hand-grenades. But recent news from the sports-meets-technology department causes us to consider adding "tennis" to the list -- and therein the controversey begins.
It turns out that "close" really does count in tennis. Just ask Wimbledon tennis champion Roger Federer, who during his final 2007 match against Rafael Nadal, claimed that Nadal hit a ball that appeared to land out of bounds, but was declared in bounds. At the center of the storm that erupted was the Hawk-Eye Officiating System , a machine-vision application that is supposed to determine without a doubt whether a ball lands inside or outside the line. In 2007, Federer said Nadal's ball landed outside, but the Hawk-Eye system said otherwise. And the umpire went with Hawk-Eye.
The way the system works, at least according to the Hawk-Eye web site, is that multiple cameras are set up around the periminter of the court. Vision processing then identifies the center of the ball within each frame of each camera. Camera movement is compensated for by also tracking the lines of the court. The system then triangulates the information from each calibrated camera to provide the 3D position of the ball. This process is repeated for each frame so that the 3D positions of the ball can be combined to produce a single trajectory of the flight of the ball. Finally, the trajectory calculates at exact bounce contact area the ball made with the court. Sounds simple enough, right?
But with the 2008 Wimbledon just weeks away, controversey over the technology has boiled up again, thanks to the paper You Cannot Be Serious! Public Understanding of Technology with Special Reference to `Hawk-Eye' written by Harry Collins and Robert Evans of the Cardiff School of Social Science challenges whether devices such as the Hawk-Eye system can actually always be right. According to the research team, the Hawk-Eye machine reported that the ball nicked the baseline by 1mm. However, Hawk-Eye Innovations, the device's manufacturer, also states that the average error of the machine is 3.6mm. If the Cardiff analysis is correct, the errors can be even larger than 3.6mm on some occasions. The International Tennis Federation, which tests the machines for use, would accept that Hawk-Eye had passed its test if it called the ball in by 1mm while the true position was out by 5mm.
In their paper, Collins and Evans introduce the "automated decision principle" which states that automated sports decision aids should not correct, but should reproduce human systematic errors -- the typical errors made by human judges and viewers such as calling a tennis ball "out" when it looks out to everyone even if the electronics suggest it might just have been in. At the same time, the devices should be used as they are now to correct or reduce human random errors, which come from lapses of concentration, an obscured view or very fast action -- but the fact that the machine can also make mistakes should always be clear. If adopted, this principle, would involve significant changes in the way devices such as Hawk-Eye are currently used in sport.
"To avoid any chance of misleading the public," says Collins, "we believe that Hawk-Eye's conclusion should be accompanied by statements and/or displays of the size of the possible errors, as is normal in science. This would ensure that the public is much better informed as to the limits and possibilities of technology."
The analysis concludes that devices like Hawk-Eye might be in danger of unnecessarily changing the traditional nature of certain games because they don't take into account how humans systematic when making judgements.
Of course, I'm wondering what the point of all this is. We all know that the other team always cheats and that umpires are uniformly blind. That's the beauty of sports.