License Plate Capture 200ft

Hello all,

I have a problem with capturing license plates at 200ft away at night. I have always used the same configuration but within 100ft. This time it is 200ft away and i cant get the plates to pop with the IR. Same 5-50 lens and same field of view.

I tried putting the IR closer to the object but still can't see it. It is at an angle where the camera has a straight shot.

Any tips that you guys can help me with? Does the IR need to be at the same viewing angle as the camera? I have IR at the camera too but it doesnt seem to go far enough.


It would help if you could provide images of what you are getting.

IPVM has some really great information that might be helpful:

http://ipvm.com/report/low_cost_license_plate_capture_shootout_2014

http://ipvm.com/report/lpc_test_2014

http://ipvm.com/report/license_plate_capture_shootout

http://ipvm.com/report/realities_of_license_plate_recognition_lpr

Beyond these resources, I'll take a stab at speculating.

Zoom: You indicate that you're using the "Same 5-50 lens and same field of view," and that you've gone from 100 ft to 200 ft distance. I'm not sure if you mean that you are zooming the lens to achieve the same absolute field of view at 200 ft as you had achieved at 100 ft, or if you mean that you have the same angular field of view at both distances. The latter would deliver 1/4 the pixel density of the former.

Illumination: If you placed the illuminator near the license plate so that the license plate itself is illuminated at the same intensity, that energy still has to travel twice as far back to your camera. Even with this optimistic assumption, you will only receive 1/4 as much energy from a plate 200 ft away as from a plate 100 ft away, and this is when the license plates are illuminated at the same brightness levels!

If the "same" field of view is angular and the illumination on the plate is NOT as bright at 200 ft as at 100 ft, then you have all of these effects working together to greatly decrease sensitivity and contrast.

On the other hand, if the same field of view means the same absolute width, and if you are illuminating the license plate with 4x the brightness at 200 ft as at 100 ft, and you have decent quality optics, then I haven't any idea. If this is the case, let's just blame it on anisoplanaticism (haha, joke). But, more likely, you've hit the Rayleigh limit of the optics and you simply can't achieve the focal plane array's pixel resolution at that zoom. If you can report the f/# of the lens at both the 100' and the 200' zoom levels, and the focal plane array area and resolution, we can make a first order assessment of whether or not that is likely the issue. Also, if that's the issue, then you will see better results in the daytime than at night. If so, then possible fixes include using better optics, or getting a focal plane array with larger pixels, or switching from infrared to blue light illumination, or all of the above.

So, I've made a lot of assumptions, and maybe they're all out to lunch (sorry!). Hopefully IPVM's really thorough information and examples will be more helpful and get you onto the right track.

Best of luck!

...switching from infrared to blue light illumination, or all of the above.

I'm intrigued about this blue light illumination, can you explain a bit more?

My first thought was that you meant UV, but that seems a bit hazardous and would seem to take a lot more power than IR.

Then, I thought you might be referring to the blue lens color (when off) of some newer GaAs LED's. But since the blue bezel is there to reduce the visible red in the emission, thereby making them more covert, I don't think that in itself would help.

Is it something new?

You're right, ultraviolet illumination should provide improved resolution compared to blue light, but many lenses lack UV transparency, and some folk do try to discourage UV exposure.

TL/DR:

In general, a given lens can focus shorter wavelengths of light to a finer point. Blue light is almost 1/2 the wavelength of the infrared illuminator's light, so blue light could provide nearly twice the resolution of infrared light, all other things being equal.

Background:

From Googling Rayleigh Diffraction Limit, or Airy Disc:

An Airy Disc, the smallest disc that a perfect lens can focus, is given by this formula:

Airy Disc spot size (um) = wavelength (um) x 2.44 x f/#

The above equation defines the diameter of a circle measured from the spot's first nulls through it's central peak. At good contrast levels, even when the Airy Disc covers 2 to 3 pixels, the brightest parts of adjacent peaks can still be resolved at the single pixel level. Because of this, then for high contrast situations, a good surrogate equation might be as follows (since 2 to 3 is about equivalent to 2.44):

equivalent pixel resolution = wavelength x f/#

Blue vs Infrared:

Blue light wavelength is about 475 nm.

The wavelength of many infrared illuminators is about 850 nm.

As a result, for a given lens, blue light can be focused almost twice as finely as can infrared illumination.

Example (with credit to Chris Dearing for the explanation):

You can often see this Rayleigh diffraction limit when using long lenses with low f/#s.

In the hotlinked test, Samsung's SNB-5004 was paired with a Fujinon 8-80 mm, f/1.4-f/14 lens. 1/3" imagers are typically on the order of 4.8 mm x 3.6 mm, with pixel size on the order of 3.65 um. Running the surrogate formula for the 80 mm focal length yields the following results for infrared and blue light:

@ Infrared: 0.850 um x f/14 = 12 um equivalent pixel resolution (about 3 pixels)

@ Blue: 0.475 um x f/14 = 6.7 um equivalent pixel resolution (about 2 pixels)

This suggests that for the referenced test, at infrared, the Rayleigh diffraction limited linear resolution can only attain about 1/4 the resolution that could otherwise be attained by the focal plane array, and at blue, only about 1/2 the resolution of the focal plane array.

Another observation is that, when mixing images of varying resolution, the eye can generally perceive features at the highest available resolution, provided the eye is not insensitive at that color range. Consider for example how FLIR ONE's superposition of a visual light image over a very low resolution thermal image gives humans the sense that they are observing the thermal image seemingly at visual resolutions.

It appears that the above theoretical results are consistent with the referenced test results. In the daytime, with white light, the observed resolution was about 1/2 of the resolution implied by the focal plane array resolution, as follows:

Max zoom at 500': geometric optics suggested ~ 38 ppf, tests yielded ~ 20 ppf.

Max zoom at 300': geometric optics suggested ~ 60 ppf, tests yielded ~ 35 ppf.

Max zoom at 100': geometric optics suggested ~ 190 ppf, tests yielded ~ 80 ppf.

Summarizing:

Both theoretical and empirical observations appear to suggest that at higher f/#s, the Rayleigh diffraction limit can lead to poorer contrast and to resolution that is substantially lower than would be implied by the resolution of the focal plane array. In such cases, better results are likely to be obtained with white light (or better still, with blue light) than with infrared light.

Horace I think that you have made an excellent theoretical case for license plate capture using blue light illumination, especially in the off-chance that we are actually diffraction limited here.

Even though one can't help wonder, considering your quick access to obscure blue light facts, if you are perhaps a spokesperson for the 1/2 micron wavelength itself? ;)

In any event, I'm wondering if motion blur won't be a problem, due mainly due drivers stomping on the brake after being bathed in blue by a bright spotlight, on an otherwise dark night. Definitely will slow them down a bit...;)

Glad to be doing my part to demonstrate yet again just how important it is to season an egghead's enthusiasm with some real world observations and experience. Uh, would you please excuse me while I run outside and disable that Super Bright 86 LED Blue Strobe?

Glad to be doing my part to demonstrate yet again just how important it is to season an egghead's enthusiasm with some real world observations and experience.

Don't let me bother you, I had just never heard of it until you brought it up out of the blue.

Looking at the pictures you may be right about the MTF, it looks like possibly a lens (or rolling shutter artifact).

As for experience, I can pretty much guarantee I will never have more than C.D. ;)

I initially did this analysis for that paragon of high quality video products, Foscam. The IR was fuzzy, but the visible was clear. A quick analysis (using the above formula) showed that the IR could never achieve the focal plane array resolution, because it was Rayleigh limited.

Since then, I've decided that it's worth checking the Rayleigh limit before you buy. You might still decide to go with it, but at least you'll be an informed buyer.

This is within 100ft with same specs and configuration F1.4 5-50 etc

This one is about 200ft away. With double the amount of IR illumination at the camera as well as another one near the target. Do i need more IR by the camera or the target?

Thanks for those images. You probably didn't need us to tell you that the second plate appeared woefully underexposed at 200' as compared to your image at 100'.

Here's the theory, which assumes that the camera is not Rayleigh limited at the 200' zoom.

If you were only adding illumination at the camera, and you wanted the camera to show the plate as brightly at 200' as it had when the plate was at 100', then you would need 16x the illumination at the camera. That's probably unaffordable, or unsightly, or both.

Let's mix it up a bit: what illumination is required to achieve equivalent brightness at the camera, depending upon where you place the illumination? Let's call the baseline the 1x case, with the camera and illumination co-located and the plate at 100' distance. Now consider the camera and plate separated by 200':

(case 1) Illumination at camera: 16x illumination required

(case 2) Illumination at 100' from plate: 4x illumination required

(case 3) Illumination at 50' from plate: 1x illumination required

From your comment, it seems that you have placed "another one near the target." I take this to mean that you are fulfilling the conditions of (case 3), plus a little bit more (mostly meaningless) illumination back at the camera, yet your experience doesn't agree with my theory. Is that correct?

...because I was going to recommend an experiment in which you move that extra illumination to within 50' of the plate, temporarily if need be, then have someone drive through the detection zone and see how it looked. I was going to say, ... "and you will see that this case with 1x illumination at 50' from the plate, with the plate 200' from the camera, should look about as bright as your baseline 100' case which had 1x illumination only at the camera."

But it seems as if you have done that, and it certainly does not match my claims. Is that correct? And you're pretty sure this illuminator is pointing at the plate, right? And the closer illumination is of comparable power to the one that was co-located with the camera to give you that acceptable plate at 100', right?

So now, I'm wondering, could you also have been using some auxillary illumination closer to the plate for that 100' baseline case? I mean, is there something I'm not understanding correctly about your setup that gave the acceptable results at 100'?

Otherwise, the last thing I'd like to verify is that the camera is not Rayleigh limited at the 200' zoom. I'd like to calculate whether or not your camera's lens is capable of resolving the numbers at that distance, regardless of the illumination.

Thanks for providing the lens info: f/1.4 5-50. Absent any other info, I'll assume you're using max zoom at 200'. The given f/1.4 would be for the 5 zoom. 50 zoom would yield f/14, at best. But at 100', to keep the same linear field of view, the lens would probably have been at f/7.

Going back to Airy Disc formula,

@ Infrared: 0.850 um x f/14 = 12 um equivalent pixel resolution

@ Infrared: 0.850 um x f/7 = 6 um equivalent pixel resolution

At 100', you resolved the plate with an effective pixel linear dimension of 6 um.

At 200', you could not resolve the plate with an effective pixel linear dimension of 12 um. 2x worse linear resolution is 4x worse area resolution. Now let's zoom in on the first & second plates to see if that sort of difference might tell the story...

The 100' plate (left) appears fairly clear and perhaps even over resolved.

The 200' plate (right) does appear more coarse, but also appears under illuminated. Looking closely, you can almost resolve some characters. For example, is the 2nd character an M? Is the 4th an R? Is the 5th a D? It's teasing us from the edge of ledgibility.

Bottom line: all this theory hasn't given us much insight. I'd say, move all of those detached illuminators up & bury that plate in brightness. If it still looks like the above comparison, Rayleigh is to blame. Otherwise, let's just blame it on anisoplanaticism.

A common practice is to use an IR beam width which matches the lens being used. For example to capture a plate at 100' I would recommend a lens HFOV be no more than 15 degrees, and to capture at 200' use 7.5degrees. Match the IR beam width to that of the lens you are using. One of the most crucial points for license plate capture is also to over power the IR light. Basically for 100' capture, use an IR that is recommended for illumination to 200'. So double light intensity for distance needed when doing LPR applications and match the light beam width to that of the lens being used and that has been proven to be very successful for my LPR applications. The obvious details as well are mounting height and degree angle away from the plate when mounting your camera. Keep height and angle to absolute minimum and results will be much more satisfying.
Perhaps it's a visual defect on my part, but if all things are equal except distance......my eyes tell me the angle is greater at 100' than 200' which would normally work in your favor except I see more direct headlight at 200'. I wonder if he drove a test car at night with headlights off what the image would appear as.