It took a computer to determine whether the ball had crossed the goal line during a French strike in the France-Honduras match at the Brazilian World Cup. Goal-line technology based on the use of 3D cameras helped the referee make a decision: the strike was allowed. Tekst: Marc Laan
This form of ICT technology is just a basic example of future artificial intelligence applications, Ronald Poppe of COMMIT/project IUALL explains. “For example, we'll be able to predict a shot on goal.”
As a university lecturer in computer vision at Utrecht University, Poppe is specialised in analysing matches on the basis of live video images. “At present, television networks still need humans to predict when and how the next goal will be scored. It's all a matter of intuition,” Ronald Poppe explains. “In some cases, the camera actually misses a goal.”
According to Poppe, it will not be long until artificial intelligence-based systems can predict a shot on goal and calculate the ball's trajectory. “The computer can then direct the cameras to that section of the pitch before the goal is actually scored.” However, Poppe also points out that current artificial intelligence technology is not quite there yet.
“That's where we're headed though,” he predicts. “Predictive technology can be extremely valuable to the networks that broadcast these matches. After all, they want the most visually attractive images they can get. At present, the directors in charge of live broadcasts often control the cameras with a joystick, based on their own intuition. Within a few years, the cameras will be self-controlled. This technology is already applied in the current generation of outdoor CCTV monitoring systems, which can recognise suspicious behaviour by car park burglars with a relatively high degree of accuracy. The very same technology can be applied one-to-one in sports matches.”
Here, the emphasis is on analysing players' movements on the pitch. “At present, we still analyse their movements manually, on the basis of video recordings. For example, researchers will count the number of steps a player takes in the run-up to a penalty shot. They tally up the winners of each header duel, and the force of each shot on goal. “We're currently developing algorithms in order to automate these analyses. The next generation of software will be able to identify individual players during live footage, track their position on the pitch and determine whether they are defenders or strikers. We're already quite far down the road in that process. The next step will be software capable of recognising more complex situations, such as players breaking through the opposing team's defence or a great one-two pass.”
World Cup audiences are already getting a taste of the fledgling computer science's first results. The lower section of the TV screen displays figures on ball possession percentages per team, the number of shots on goal and the number of goals. “Intelligent computer algorithms are extracting this information from the live video feed.” Another familiar application is the use of heatmaps: colourful graphs that highlight the parts of the pitch where a player has spent most of the match.
Goal-line technology is already widely accepted in the sports of field hockey, tennis and football. Poppe: “Current technology is already extremely effective in terms of determining whether the ball has crossed over the pitch boundary or gone in the goal. The next challenge will be to identify offside offences. If these efforts are successful, they might eliminate the need for a third and fourth referee. However, this would require computer systems capable of tracing both the ball and the players. This will require mathematical interpretation which must be entirely accurate in order for the technology to be accepted.”
“Eventually, we'll obviously want systems that are capable of recognising a dive. Was the striker really tackled by the defender in the penalty area, or is he just acting? In the end, such a system might result in fewer penalties, which could make for less exciting and emotionally-charged matches. Thankfully, it will be years before we have systems capable of accurately interpreting players' posture and behaviour. I don't think we can expect to see a robot referee within the next ten years.”
Ronald Poppe has also found other applications for the online behaviour recognition technology. “We're also active in the field of ambient intelligence. For example, we monitor the behaviour of elderly care home residents suffering from Alzheimer's disease as part of a pilot study. The software can recognise patients wandering around aimlessly. The system can then automatically give off a soothing stimulus, such as a song from the patient's youth.”
Poppe has applied the same technology to children playing tag. The artificial intelligence system recognises the patterns in their movement. “For example, we can activate the children by projecting colourful circles on the playing floor to allocate them a new role in the game.”. Another pilot project: Both research projects are funded by COMMIT/.