Vehicle: Car: Speed: Control
Double camera solves traffic tracking   (0)  [vote for, against]
Use two low-cost low resulotion cameras to get high speed resolution

Warning: idea for technology geeks.

A problem when making speed tracking systems (which is what my company is currently doing) is the high cost of high speed high resolution cameras. What I need is only to catch the car every half meter so I can track the car and it's next position (at most 50 images/sec) but once a new car enters the range I want a very high speed (every cm counts) camera for recording the cars' speed.

Solution: Use two low speed cameras. The second takes a picture at a very close interval (even 1 ms away) which could easily be achieved if both cameras are controlled by one source (PC and image card).

This could lower the cost of cameras to less than $100, and if they are USB cameras, nothing more besides PC and software!
-- pashute, Nov 04 2002

Hope you didn't sign an NDA.
-- phoenix, Nov 04 2002


Hehe.
-- bristolz, Nov 04 2002


Confused by interchanging resolution and speed: not the same thing when I last looked.
-- DrCurry, Nov 04 2002


but tend to go together in price ranges.
-- pashute, Nov 04 2002


Yeah, but which are you really after here? (And I'm still not clear on what the half-baked idea here actually is.)
-- DrCurry, Nov 04 2002


The essential idea here is to measure the speed of an object by taking two pictures of it 1ms apart; using two cameras to accomplish this may be cheaper than using one camera (which would have to be able to take pictures in 1/1000 second).

While I appreciate the notion, things aren't quite so simple as you suggest: since the two cameras most likely won't be in exactly the same place(*), the two pictures will be taken from slightly different angles. While this may be dealt with in various ways, it adds more complexity than IMHO it's worth.
(*) It would be possible to use a half-silvered mirror to split the image going to the two cameras, so as to allow them to photograph from exactly the same angle, but this too would add complexity.

One thing I must admit, though: I'm not quite clear why you want two pictures so close together? The closer together the two pictures are taken, the more resolution is necessary to resolve the vehicle's speed. A car which is travelling at 180km/h will traverse fifty meters/second. In 1/50 second, therefore, it will traverse one meter. If your camera has 1cm resolution, then, comparison of the two pictures will allow speed measurement accurate to 1.8km/h. If the two pictures were only 1ms apart, however, then a car travelling 180km/h would only traverse 5cm between the two pictures. A 1cm camera resolution would only measure speed accurate to 36km/h.
-- supercat, Nov 04 2002


I suspect its partly just a cultural thing creating some 'signal-to-noise' issues.

He also tends to describe some of the specific points of the idea, without describing the general idea itself or what it's for. I've been able to interpolate them for the most part, given some engineering clues here and there, but not everybody has that background.

Pashute, you'll discover that success, voting-wise, in the bakery is as much about the people there and their general presentation likes / dislikes as it is about the merits of the ideas themselves. It's an excellent study in group behavioral psychology.
-- RayfordSteele, Nov 05 2002


Yeah, man, get yer tape measure out!
-- Nick@Nite, Nov 05 2002


Maybe just a really, really sensitive camera that measures blue or red shift.
-- bristolz, Nov 05 2002


or we could make the troopers do their jobs....
-- tkeyser, Nov 05 2002


One reason I see that they use photos for speed is with red-light cameras here in Aus. They need to take two photos anyway, the idea being that if you just roll across the stopping line and then stop, this will show in the second picture and you don't get charged for completely running the light. If you are charged for running the light, they measure the distance you have traveled in the time between the two photos and determine your speed from that. The cameras have 4 lenses (like passport photo cameras) so that the multiple pictures can be on one negative. By taking the Upper-Left and Lower-Right at first and then taking the Upper-Right and Lower-Left pictures, the problem of differing lens position is accounted for.

However, like supercat said, the interval is approximately 1/25 to 1/50 second between shots, so there is a significant distance covered in the meantime.
-- reap, Nov 06 2002


[bristolz] - wouldn't work - people in blue cars would never get speed tickets. Far better to use ultra-sensitive weighing scales in the road to measure the increase in the weight of the car caused by travelling at high speed, as predicted by General Relativity.
-- hippo, Nov 06 2002


The idea is sound. But isn't a little overly complicated when two induction coils set a few meters apart could do the same thing? Or two airhoses. Or two IR beams.

All those systems are relatively cheap and use proven technology. I can't see why these methods would need to be altered in any way. But, as your company is developing this system, there must be some sort of demand for it. Kudos to you for coming up with a way to make an unneccesarily complicated system a bit less complex.

+
-- rapid transit, May 24 2003


Both speed and red-light cameras linked to citation-writing computers are just taxation in the guise of public safety. This idea should work in principle (altho a cheap, low-speed camera will likely take blurry photos of speeding cars), but it gets a fishbone, also as a matter of principle.
-- ConsultingDetective, Jan 31 2004


There was a time awhile ago when I probably would have gotten nailed by a traffic light camera had one existed, but my course of action was necessary to prevent a very severe accident. The intersection had two 'ahead' lanes and one left-turn-only lane. The intersection turned yellow when I was some distance away, and I was stopping at it just as the light was about to turn red; there were cars in both of the other two lanes. As I was stopping, however, I noticed a vehicle behind me approaching at very high speed. Recognizing that the vehicle could not possibly stop in time to avoid a collision, and had no open lane to swerve into, I floord the gas just as the light turned red. I was subsequently passed within the intersection by the bat-out-of-hell.

As it was, I probably crossed the threshhold very shortly after the light turned red, and significantly before before the cross traffic turned green. Had I not done so, I would have entered the intersection a second or so later when I got clobbered at 60+mph. Indeed, I was not expecting to avoid a collision entirely--I was merely hoping to reduce its severity by reducing the difference between by vehicle's speed and the bat-out-of-hell's.

Had there been a camera at that intersection, should I have gotten a ticket?
-- supercat, Jan 31 2004


Is the Heisenberg Uncertainty Principle a valid defense for speeding tickets ?
-- normzone, Jan 31 2004


Almost two years since posting. First let me say, the idea works, and is now being used.

NDA: My company, my idea, my product. This was a small side issue.

The need was to replace the very expensive to install and maintain coils for speed control of various types, such as traffic light by speed etc. High speed (and much less so resolution) was what I was after, because a car going at 180km/h = 50meters/sec with a regular low cost (half second per pic?) camera, would see the car in its frame, and next pic would be totally missing, or perhaps wouldnt even catch the car if say a span of 40 meters of road is being shot from above. The whole scene would show empty, while the car had a chance to enter and pass the whole part.

Supercat's "not so simple" problems are actually quite trivial in image processing, especially using ready-made toolkits like National Instruments IMAQ etc.
-- pashute, Jun 22 2004



random, halfbakery