Hoe does the new ESPN Shot Spot work?
January 28, 2007 11:53 AM Subscribe
How does the ESPN Shot Spot work? I've been watching the Australian Open with some skepticism. Why is everyone so confident that the calls are accurate?
If you haven't seen it, Shot Spot is a computer generated graphic that shows exactly where the ball lands on court. It now decides calls that are challenged like NFL instant replay. But it's not video footage, so how does it work? I know it has something to do with eight cameras, but can't find anything more on it on the web.
If you haven't seen it, Shot Spot is a computer generated graphic that shows exactly where the ball lands on court. It now decides calls that are challenged like NFL instant replay. But it's not video footage, so how does it work? I know it has something to do with eight cameras, but can't find anything more on it on the web.
It's actually pretty simple: multiple sensors and camera angles allow the ball's position and velocity to be captured and these are then used (along with wind speed/direction etc.) to calculate a probable trajectory.
I remember seeing a demonstration of the HawkEye system (I think it was during rain at Wimbledon) and it was incredibly accurate. Channel 4 used to use it during the cricket during contentious LBW decisions.
posted by alby at 12:53 PM on January 28, 2007
I remember seeing a demonstration of the HawkEye system (I think it was during rain at Wimbledon) and it was incredibly accurate. Channel 4 used to use it during the cricket during contentious LBW decisions.
posted by alby at 12:53 PM on January 28, 2007
Response by poster: Jourman2, thank you for the articles, they are somewhat informational, but we are still not getting to the bottom of this.
All the web shows is 6 cameras shooting 60 frames a second, that feed a computer, which can handle billions of equations, which form a graphic.
But how do six camera angles, basically just pixel footage, become a fully accurate vector. How does the computer read the footage to single out the ball? Can it trace yellow?
posted by elfollador at 5:59 PM on January 28, 2007
All the web shows is 6 cameras shooting 60 frames a second, that feed a computer, which can handle billions of equations, which form a graphic.
But how do six camera angles, basically just pixel footage, become a fully accurate vector. How does the computer read the footage to single out the ball? Can it trace yellow?
posted by elfollador at 5:59 PM on January 28, 2007
According to the Wikipedia entry for Spot Shot, ESPN has not released the details of how this works. However, it is probably similar to the first down line marker and the FoxTrax puck effects that have been used in Football and Hockey, respectively.
It's not just camera angles. The cameras direction and zoom level are also factored in. This is how they do the first down line in the NFL. By knowing where exactly the camera is (i.e. 100 feet away and 30 feet above the playing field), what direction it's facing (12 degrees right of midfield), and how far in it is zoomed, you can use mathematics to determine where on the TV screen the 30 yard line will be, and superimpose a line over it.
It is likely, then, that Spot Shot uses all of these camera angles and mathematics, along with detecting difference in color (tennis ball vs. court) you can get a fairly accurate guess of where in 3 dimensional space the ball is even with just one camera view. However, to reduce margin of error, 6+ camera views seems like the logical thing to do.
Again just theory -- but using audio cues or having an accelerometer in the ball would allow detection of when it hit the ground -- presto, now you know at what moment in time to check the frame in all 6 camera angles, and can calculate with pretty high accuracy where the ball actually hit.
posted by twiggy at 6:22 PM on January 28, 2007
It's not just camera angles. The cameras direction and zoom level are also factored in. This is how they do the first down line in the NFL. By knowing where exactly the camera is (i.e. 100 feet away and 30 feet above the playing field), what direction it's facing (12 degrees right of midfield), and how far in it is zoomed, you can use mathematics to determine where on the TV screen the 30 yard line will be, and superimpose a line over it.
It is likely, then, that Spot Shot uses all of these camera angles and mathematics, along with detecting difference in color (tennis ball vs. court) you can get a fairly accurate guess of where in 3 dimensional space the ball is even with just one camera view. However, to reduce margin of error, 6+ camera views seems like the logical thing to do.
Again just theory -- but using audio cues or having an accelerometer in the ball would allow detection of when it hit the ground -- presto, now you know at what moment in time to check the frame in all 6 camera angles, and can calculate with pretty high accuracy where the ball actually hit.
posted by twiggy at 6:22 PM on January 28, 2007
Just stopping by to note that I've corrected the Wikipedia article. The invaluable camera system is Shot Spot. Spot Shot is a carpet stain remover, apparently made by the fine folks at WD40.
posted by diddlegnome at 9:17 PM on January 28, 2007
posted by diddlegnome at 9:17 PM on January 28, 2007
The stupid puck thing actually used a special puck, one of the reasons it was hated.
posted by Mitheral at 9:23 PM on January 28, 2007
posted by Mitheral at 9:23 PM on January 28, 2007
The stupid puck thing actually used a special puck, one of the reasons it was hated.
Actually it used a regular puck, not a specially manufactured one. They took regular pucks, sliced them in half, inserted the electronics and epoxied them back together. The process added only 1/100th of 1 gram to the weight, so I find it hard to believe the special pucks caused the players any annoyance.
I think the reason it was hated because it was too videogame-ish.
Anyhow - as far as further on-topic information: One conceivable idea would be:
1 camera facing straight down, mounted directly over dead center court.
4 cameras at ground level, placed at the sides of the court, facing inward, exactly halfway between the baseline and the net.
The 4 ground level cameras could get you lateral velocity and also tell you when the ball hits the ground since they're at ground level. The overhead, then, freeze-framed at the time the ball hits the ground, could tell you where the ball hit the ground. Perhaps because of viewing angles you'd need more than one camera mounted overhead for this - perhaps 2 or 4 would be sufficient to have a more accurate picture.
posted by twiggy at 3:30 AM on January 29, 2007
Actually it used a regular puck, not a specially manufactured one. They took regular pucks, sliced them in half, inserted the electronics and epoxied them back together. The process added only 1/100th of 1 gram to the weight, so I find it hard to believe the special pucks caused the players any annoyance.
I think the reason it was hated because it was too videogame-ish.
Anyhow - as far as further on-topic information: One conceivable idea would be:
1 camera facing straight down, mounted directly over dead center court.
4 cameras at ground level, placed at the sides of the court, facing inward, exactly halfway between the baseline and the net.
The 4 ground level cameras could get you lateral velocity and also tell you when the ball hits the ground since they're at ground level. The overhead, then, freeze-framed at the time the ball hits the ground, could tell you where the ball hit the ground. Perhaps because of viewing angles you'd need more than one camera mounted overhead for this - perhaps 2 or 4 would be sufficient to have a more accurate picture.
posted by twiggy at 3:30 AM on January 29, 2007
There is a bit more to it than that. Even if the mass is the same some of the puck was carved away to fit the electronics and didn't play like a regular puck.
"From its first use, players complained that the FoxTrax puck did not move the same way a normal puck did. The FoxTrax puck also did not hold the cold as well. FoxTrax pucks became bouncy much more quickly than their regulation counterparts. "
posted by Mitheral at 7:38 AM on January 29, 2007
"From its first use, players complained that the FoxTrax puck did not move the same way a normal puck did. The FoxTrax puck also did not hold the cold as well. FoxTrax pucks became bouncy much more quickly than their regulation counterparts. "
posted by Mitheral at 7:38 AM on January 29, 2007
This thread is closed to new comments.
From what I can gather it uses high speed cameras to calculate velocity, compression, and some other stats to generate a landing point thats accurate within 3 mm.
posted by jourman2 at 12:02 PM on January 28, 2007