In 1999 research began at Roke Manor Research Ltd., a company with over thirty years of vision processing expertise. Led by Dr. Paul Hawkins and funded by The Television Corporation, the concept of “Hawk-Eye” was born. First developed in 1999, Hawk-Eye Innovations Ltd continues to grow and develop with Hawk-Eye recently creating a system for the BBC’s snooker coverage and is currently developing a goal-line football system for the English FA. The possibilities are endless….
We spoke to Paul Hawkins regarding Hawk-Eye innovations in the world of Cricket........................
How does Hawk-Eye work?
Hawk-Eye uses 6 high speed specialist vision processing cameras which are positioned around the ground and calibrated. In addition the system uses the two “Mat” broadcast cameras and calibrates them so that the graphic is always overlaid in the right place. All cameras have “anti wobble” software to deal with camera movement.
When a ball is bowled, the system is able to automatically find the ball within each
frame of video from each camera. By combining the positions of the ball in each of the cameras, the 3d position of the ball is measured through-out the delivery. The delivery is broken down into two independent components: delivery to bounce and bounce to impact. No intrinsic characteristics of the pitch are required as part of the system apart from deliveries which hit the batsman on the full. By measuring the position of the ball at multiple frames post bounce, the speed, direction of travel, swing and dip can be calculated for that specific delivery. Using these measured characteristics of the ball’s flight up to impact, the future path is predicted.
Is the Hawk-Eye system fully automatic?
3 Hawk-Eye staff are required to operate the Hawk-Eye system. Their roles include:
1. Lining up and calibrating the cameras
2. Measuring the pitch and the stumps which do vary from ground to ground
3. 1 member of staff is responsible for the virtual reality graphics and offers LBW replays and all the other Hawk-Eye features to the TV director.
4. The other 2 members of staff both are responsible for the tracking. They work independently of each other to provide redundancy, but are able to see a comparison of the two tracks. If they are different for any reason, they can be pro-active in working out why rather than being re-active after a LBW appeal. On a ball by ball basis they would do the following:
a) hit a button to tell the system that a ball has been bowled and trigger the tracking.
b) manually fine tune the point on the trajectory where interception with the batsman was made. Automatically the system is only able to determine the interception point to the nearest frame of Hawk-Eye video running at 106 frames per second. This can be improved manually and is the only way to ensure that the interception point is accurate to 5mm.
c) Tune settings to account for varying light conditions.
d) Tune settings to deal with camera wobble
How does the system account for variable bounce and movement off the pitch?
This is probably the single biggest area of the system which is mis-understood and inaccurately portrayed in the media.
The Hawk-Eye system does not use and generic characteristics of the pitch to predict the amount of bounce or sideways movement. Just as a good umpire should, each ball is measured individually. For LBW it is only the track from pitching to impact which is of relevance. By measuring the actual path of the ball up to the point it has hit the batsman, the position, velocity and accelerations on that particular ball can be determined. Figures 3, 4 and 5 provide good examples of variable bounce. The delivery to North pitches at a very similar length to the Clarke delivery, but the North delivery rose more sharply off the pitch.
What if there is a small distance between pitching and hitting the batsman?
As was stated in the performance criteria, the accuracy of the prediction is degraded if there is a small amount of data post bounce, just as it is for an umpire. However, this is a smaller problem than may be at first thought for the following reasons:
• Less post bounce data is required to predict the line than is required for the height. Height is not typically relevant for full balls where there is little post bounce data.
• If the batsman is hit on the full, then the prediction is easy, as the laws state that the umpire should assume that the ball goes straight on.
• For all of the LBW appeals in the matches where the referral system has been in place, there has been enough post bounce data to make an accurate prediction. On average there has been 3.58 meters of travel between pitching and hitting the batsman. The minimum amount of travel was an Anderson. Yorker on day 2 of the Centurion Test which had 35 cm of travel. Hawk-Eye predicted that delivery would hit the stumps half way up.
• In all instances the prediction accuracy is higher than a human umpire could typically achieve for that type of delivery.
How does the accuracy of the prediction degrade the further is has to predict?
So long as there is enough data available out of the bounce, the prediction accuracy degrades very slowly the further it has to predict, as is shown in the examples below.
The main reason for the “over 2.5 meters” element of the protocol is because it fits more naturally with the way the game has been played. If a batsman comes down the wicket, but doesn’t get to the pitch of the ball and pads up, he would typically not be given out LBW, and this element of the protocol ensures that this continues to be the case.
If Hawk-Eye is accurate, why are there instances when many people all agree that the ball did something different to what Hawk-Eye shows?
It is human nature to question things which one cannot see for yourself. Any illusionist plays on the fact that that the human brain places a great deal of emphasis on what you believe you have seen, so it becomes hard to accept that this is not reality. There have been many occasions when I have been watching cricket at home and seen a Hawk-Eye replay and felt it was wrong. I speak to my staff on site, ask them to check it, confirm that they were happy with what went to TV and to send me further data. By looking at it in more detail, I have learnt that the Hawk-Eye track was right.
Recently, other independent people have received this data, and their confidence in the system has increased as a result. This process has enabled us to build up an appreciation for the number of ways in which a broadcast replay can be deceptive.
For all official series we have offered this additional evidence to the teams, umpires and commentators. If a commentators claims that a Hawk-Eye replay is wrong in a series where the system is being used officially, and the commentator is subsequently proven to be correct, Hawk-Eye will contribute £10,000 to a charity of the commentators choice. If the commentator is wrong, we ask that they contribute £1,000 to the Lords Taverner’s!