Taxi Drivers

[flango]

Sorry but I also very much doubt that these mass produced small cameras can get anwhere near that level of detail given the confines of the optics.

The police would like us to believe such a thing I'm sure, but suggesting that it could read the registration mark from a window etching of a fast moving car from 20+ meters away is pure fantasy (or more likely- spin) imho.

Nope it's absolute fact I have seen them do it and if anyone going on the South Yorkshire Police visit wants proof I'll get them to show it to you as well. We are talking images from Fixed Post intallations (Gatso, Truvello, Specs). I would agree if you are talking about mobile speed vans where you have a teletraffic set up on something like an LTi 20-20 laser gun as the video in this case is not that high a resolution, just enough to capture the number plate and image of the driver if needed. But if you are caught on a Fixed Post Installation the resolution is incredibly high (will try and find out the exact resolution)

 

[flango]

Speed camera spec

Apparently Gatso cameras around this neck of the woods have the following spec.

Lens Schneider Tele-Xenar f3.8/90mm
Angle of view 23.4 degrees horizontal, 15.6 degrees vertical
Distance setting 20m fixed

Automatic Diaphragm
Film sensitivity setting 25 - 1600 ASA (ISO)

Resolution 6 Megapixel

Although not sure of the actual optics of the Truvelo, resolution is the same at 6 Megapixel.

Permanent Specs camera's (i.e not used as a temporary measure in roadworks) operate at 7.2 megapixel.

Normal video camera for mobile vans is

JVC KY-F550 A high resolution three chip CCD Colour Camera for high definition and picture quality. This camera provides 800 lines of Broadcast quality digital real time video at 25 frames of 50 fields per second. It also has the capability to operate in low light conditions down to as low as 1 lux. (Full spec HERE)


With image enhancement software and equipment, you could spot a pimple on a flea's ar*e on your winscreen at the resolution of fixed post cameras.

 

[jonm]

So if we assume that the area of the image is 2M by 3M, that gives 6M sq MM and 1 pixel per Sq MM. So assuming the lens is capable of giving a sharp image the electrickey appears capable of giving a clear image of a VIN no.

food for thought

 

[BTB 500]

So if we assume that the area of the image is 2M by 3M

Did you work that out? Haven't got time to do the maths now, but that's the key (derive from the angle of view and distance).

 

[Dryce]

Apparently Gatso cameras around this neck of the woods have the following spec.

Lens Schneider Tele-Xenar f3.8/90mm
Angle of view 23.4 degrees horizontal, 15.6 degrees vertical
Distance setting 20m fixed

Automatic Diaphragm
Film sensitivity setting 25 - 1600 ASA (ISO)

Resolution 6 Megapixel

Although not sure of the actual optics of the Truvelo, resolution is the same at 6 Megapixel.

At 10m this is (I think) about 4m wide view at 23 degrees.

So 6Mp on a 3:2 sensor would give you 3000 x 2000 pixels.

So 4 metres divided by 3000 is 1.3 pixels per mm.

Double that for 20m distance and that's 0.65 pixels per mm.

With image enhancement software and equipment, you could spot a pimple on a flea's ar*e on your winscreen at the resolution of fixed post cameras.

So not quite. You might spot a pimple on somebody's face.

But you won't spot the flea let alone it's ar*e or the pimple set there upon.

 

[flango]

So not quite. You might spot a pimple on somebody's face.

But you won't spot the flea let alone it's ar*e or the pimple set there upon.

You will with the image enhancing software and equipment they use as I posted earlier. I'm not talking about the raw image but the enhanced one they use to read the information. My point was that the basic quality of the raw image is a lot better than people think and can be enhanced to a great degree of detail

 

[Dryce]

You will with the image enhancing software and equipment they use as I posted earlier. I'm not talking about the raw image but the enhanced one they use to read the information. My point was that the basic quality of the raw image is a lot better than people think and can be enhanced to a great degree of detail

It can't be any better than the input signal - which is down to the pixels and then beyond that their sensitivity to colour and light/dark. (I'm assuming the images are monochrome?) So if the pixels and lens limit you to say 1mm per pixel it will be worse than that in reality because you'll need several pixels in each axis to discriminate features.

If image enhancing software is used to try and beat these limits then that causes problems as regards evidence. Because it is effectively guessing below this point. That doesn't mean that it's not effective at producing good data/images - just that it's working on probabilities. So you might generate intelligence from it as long as you understood its limitations.

And there is subject movement which is why speed cameras tend to use flash - to freeze the image more precisely. A vehicle at 30mph travels 13300 mm per second. So a flash of 1/10000 would mean 1.3mm movement - ie. 1 pixel at 10m, 2 pixels at 20m. At 1/50000 then this would be more like 0.25mm of blur.

The motion blur will increase proportionate to the speed of the vehicle and decrease proportionate to the flash duration.

[ I have no idea what the flash duration is likely to be on a speed camera setup.]

 

[Deleted member 6183]

It can't be any better than the input signal - which is down to the pixels and then beyond that their sensitivity to colour and light/dark. (I'm assuming the images are monochrome?) So if the pixels and lens limit you to say 1mm per pixel it will be worse than that in reality because you'll need several pixels in each axis to discriminate features.

If image enhancing software is used to try and beat these limits then that causes problems as regards evidence. Because it is effectively guessing below this point. That doesn't mean that it's not effective at producing good data/images - just that it's working on probabilities. So you might generate intelligence from it as long as you understood its limitations.

And there is subject movement which is why speed cameras tend to use flash - to freeze the image more precisely. A vehicle at 30mph travels 13300 mm per second. So a flash of 1/10000 would mean 1.3mm movement - ie. 1 pixel at 10m, 2 pixels at 20m. At 1/50000 then this would be more like 0.25mm of blur.

The motion blur will increase proportionate to the speed of the vehicle and decrease proportionate to the flash duration.

[ I have no idea what the flash duration is likely to be on a speed camera setup.]

So is that a Yes or a No then?

 

[flango]

It can't be any better than the input signal - which is down to the pixels and then beyond that their sensitivity to colour and light/dark. (I'm assuming the images are monochrome?) So if the pixels and lens limit you to say 1mm per pixel it will be worse than that in reality because you'll need several pixels in each axis to discriminate features.

No Images are full colour

If image enhancing software is used to try and beat these limits then that causes problems as regards evidence. Because it is effectively guessing below this point. That doesn't mean that it's not effective at producing good data/images - just that it's working on probabilities. So you might generate intelligence from it as long as you understood its limitations.

Image enhancement by recognised techniques is now admissable as eveidence and was used in the 2 cases I illustrated.

And there is subject movement which is why speed cameras tend to use flash - to freeze the image more precisely. A vehicle at 30mph travels 13300 mm per second. So a flash of 1/10000 would mean 1.3mm movement - ie. 1 pixel at 10m, 2 pixels at 20m. At 1/50000 then this would be more like 0.25mm of blur.

No flash on SPECS cameras @ 7.2 MP just infra red for night vision.

The motion blur will increase proportionate to the speed of the vehicle and decrease proportionate to the flash duration.

Not sure of flash duration but is synchronised with shutter speed at 1/1000th and takes 2 pictures per second

[ I have no idea what the flash duration is likely to be on a speed camera setup.]

I have seen some of these enhanced Images and I'll get them to display some on the South Yorks Police visit on some of them you can clearly see what make of watch the driver is wearing. I'm not into photography and would not know where to start but I do know it is possible

 

[alexander patie]

So 6Mp on a 3:2 sensor would give you 3000 x 2000 pixels.

So 4 metres divided by 3000 is 1.3 pixels per mm.

Double that for 20m distance and that's 0.65 pixels per mm.

sorry to nit pick, but 4m/3000 pixels = 1.3 mm per pixel.

doubling that gives us 2.6mm per pixel at 20m - which equates to around 0.385 pixels per linear mm, or 0.148 pixels per mm2.

add on the angle of the object and any blurring as has been said and suddenly reading that windscreen vin becomes tricky - probably right at the limit of the image. i am not trying to disprove what is being said here though - i do believe it is possible, but doubt that it is always legible. nor do i expect any conclusions about the 'mooner's' prostate can be drawn.

 

[Dryce]

sorry to nit pick, but 4m/3000 pixels = 1.3 mm per pixel.

Valid nitpick.

 

[BTB 500]

Well ... I've just checked some good quality (ISO 80-100) full-frame frontal pics. of members cars taken at GTGs with an 8MP DSLR (Canon EOS 20D) and 12 MP compact (Canon G9), and there is no way you'd be able to get a VIN from any of those. The expiry year/month on the tax disc is OK, but nothing much smaller than that is legible ... no matter what processing you do to the image.

So doing it with a 6MP image of a moving car taken from 65 feet away would be impressive. Possibly shooting downwards (from a higher vantage point) would help a bit - not such an acute angle through the windscreen glass

 

[flango]

Well ... I've just checked some good quality (ISO 80-100) full-frame frontal pics. of members cars taken at GTGs with an 8MP DSLR (Canon EOS 20D) and 12 MP compact (Canon G9), and there is no way you'd be able to get a VIN from any of those. The expiry year/month on the tax disc is OK, but nothing much smaller than that is legible ... no matter what processing you do to the image.

So doing it with a 6MP image of a moving car taken from 65 feet away would be impressive. Possibly shooting downwards (from a higher vantage point) would help a bit - not such an acute angle through the windscreen glass

Sorry not being funny but you have missed the point I was making this is only possible with Image enhancement software and equipment to get the VIN or detail and not from the original speed camera image, hope that clarifies.

 

[Dryce]

hope that clarifies.

Nope. Niet. Nein.

But looking forward to finding out more about this.

 

[BTB 500]

Sorry not being funny but you have missed the point I was making this is only possible with Image enhancement software and equipment to get the VIN or detail and not from the original speed camera image, hope that clarifies.

You can't enhance something that isn't there though. Here's an example, this is from a RAW format 8MP DSLR image - ISO 100, 1/100 sec. at F9

You can see the individual pixels ... there simply isn't any more detail than that.

Without having one of their images to play with I'm not categorically stating that its impossible. But as an experienced photographer I find it very hard to see how it could be done.

 

[andy_k]

a couple more points

the pic of the guy mooning was not taken by a camera on any sort of raised gantry - it's looking straight through the car so it must have been taken at 'car level'

The VIN would have been at least partly obscured by the parked windscreen wiper

 

[andy_k]

here's the 'real' story

 

[flango]

a couple more points

the pic of the guy mooning was not taken by a camera on any sort of raised gantry - it's looking straight through the car so it must have been taken at 'car level'

The VIN would have been at least partly obscured by the parked windscreen wiper

Yes it was I cropped the picture from the wide camera views before I posted them. I'm not sure where the VIN is on an X5 but as they are not actually built by BMW if its the same as the others that come out of that factory in Autria then the VIN could be on the drivers side.

Sorry misread your post above refers to bike pic oops

Anyway we'll dispel all these urban myths when we Visit South Yorks Police. I'll get them to show some so there is more than me on here who will have seen them

Nice find on the original story though, these were some pics I just dug out of my archive that people have sent me relating to road traffic incidents over the years

 

[andy_k]

Well in the case of the biker they tracked him through the camera's until he reached one of the camera's on the motorway cctv network and got him by his back plate.

I'm afraid your information is wrong here as well

what happened

So, it wasn't camera cleverness that caught him either, just good old fashioned policing

 

[nick mercedes]

I'm afraid your information is wrong here as well

what happened

So, it wasn't camera cleverness that caught him either, just good old fashioned policing

It's a shame they don't devote so much manpower and investigating to most other reported crime....