SoylentNews is people
SoylentNews
from the is-a-tenth-of-a-millisecond-faster-than-the-blink-of-an-eye dept.
Velonews reports that a recent Tour de France stage was won by 5mm,
http://www.velonews.com/2017/07/tour-de-france/figure-won-stage-7_443075 which they convert to 0.0003 seconds. The article then goes on to describe the finish line camera:
The judges use a camera placed on the finish line that shoots at 10,000 frames per second. This allows them to confidently pick a winner even when differences are far too small to spot with the naked eye.
...
The cameras don’t work like a normal video camera though. They work more like a scanner.Rather than shoot frames that are thousands of pixels wide using some sort of shutter and digital sensor (the modern replacement for film), the finish line camera is a slit camera. Old slit cameras run film behind a lens. In the timing camera’s case, the design exposes a digital sensor.
A flatbed scanner is a type of slit camera. So imagine pointing one of those at the finish line and scanning the riders coming across. Frame rates can be so high because there is no shutter to close and the cameras only record a one-pixel wide image at a time (10,000 times per second). This type of camera, pointed at a finish line, is guaranteed to show you who or what got to that finish line first, because it shows almost every moment. This is also the source of the distortion we associate with finish line photos. The scanner has a set speed, and anything going slower gets elongated — anything faster gets squished.
No shutter means nothing is missed (because shutters close, and you miss that part). That’s good when the riders are crossing the line .0003 seconds apart from each other.
Anyone know about this technology? Somehow the explanation above doesn't seem all that clear.
[Ed. addition] Maybe one of these recommendations by mrpg might help?
https://gearpatrol.com/2016/07/21/tour-de-france-timed/
https://cyclingtips.com/2012/06/how-time-gaps-are-calculated/
(Score: 2) by FatPhil on Sunday July 09 2017, @02:50PM (1 child)
First we need something to roll to be the front bike wheel, perhaps a patterned plate, something which clearly looks different at different angles of rotation (so no /The Who/ or RAF target insignia).
Next, we need the beam/slit - two straight edged bits of cardboard that you can prop upright, or 2 narrow (flat) boxes like chocolate boxes if propping card is too hard. Place these near each other so there's only a narrow gap between them
Start with the wheel completely occluded behind one of the bits of the card - no wheel/plate visible.
Roll the plate a tiny bit forward - still nothing visible, roll it more - still nothing, until...
Roll the plate forward a bit until the very front edge appears behind the gap - congratulations, that plate cyclist has won - the camera will detect that front edge.
Roll a tiny bit more, notice which bit of the pattern on the plate is visible - the camera will capture that pattern as the next column
Roll a bit more, notice the pattern again - again, the camera's capturing that as the next column
Continue until bored.
It's probably best to try and draw a pencil line down the slit with each unit of time that passes - you'll be able to see how the camera captures a distorted view of the plate as it sweeps out lines at different orientations.
If you had a spokey pattern on your plate, you should notice that the spokes were all bent over time. Both forwards and backwards - basically away from the ground. See the first photo on the news site - look at those spokes.
I did a *very* quick **very very** crappy simulation with a beermat, a chopstick for a finishing line, very poor temporal resolution, and a terribly nasal voice:
http://theanna.org/tmp/wheel.mp4
I then upsampled by eye, and mapped the captured slices onto a time axis:
http://theanna.org/tmp/wheely.jpg
Hopefully the utter terribleness of that will inspire someone else to do a better one.
Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
(Score: 0) by Anonymous Coward on Sunday July 09 2017, @05:12PM
Nice sketch (the video wouldn't play here). But you had the advantage of stopping the wheel when you plotted the coordinates on each "time-line". In reality, isn't the line scanner (CCD?) serial? If so, then stuff at one end of the scanned line will be sampled before stuff at the other end???
Or maybe this line scanner really is parallel, with a clock input, and an A-D (or other digitizing) for each pixel?
The other thing that puzzles me is how they get enough light -- normally to do 10K frames/second it takes huge banks of lights--I'm thinking of the high speed video used to film car crash tests (in a crash lab).