Capturing License Plates 500 Feet Away?

JH
John Honovich
Mar 23, 2013
IPVM

A member wants to capture license plates 500 feet away (just capture, not automatic recognition). I have some theoretical concerns but wanted to see if anyone has field experience with such a project.

You are going to need a very long lens. Assuming you want a tight foot wide FoV (10', 20'?), you will need a 150mm or longer lens. With such a long lens, I am concerned about high F stops (increasing low light problems), image stabilization (with even modest shaking) and limited depth of field.

Moreover, unless there is a light source near and aimed at the car, I am concerned that headlights are going to overwhelm the camera, making it impossible to capture the license plate.

Any thoughts? Any experiences?

[Background reference: Our LPC test results from a short range scenario.]

Avatar
Brian Karas
Mar 23, 2013
Pelican Zero

Day? Night? 24/7? What state(s)? Can illumination be placed closer to the vehicles? What will the camera angle be relative to the plates? Why does it need to be so far away? What's the budget? Are you sure this isn't just a slightly less obvious version of Matt's "I need a cctv thing" thread? ;)

JH
John Honovich
Mar 23, 2013
IPVM

Day and night. Let's assume front plate and that the vertical angle will be relatively minimal, since over such a distance even a camera mounted on top of a few story building won't be that steep.

No, it's a real case and, for this, it needs to be that far away.

Avatar
Sagy Amit
Mar 24, 2013

I recently used a 1080P box camera with 80mm lens on the roof of a power plant, looking at a vehicle gate. The intention was never to read a license plate, but during the day with perfect visibility we are just 10mm or so short of that.

The camera is mounted at about 90Ft and looking at a distance of 900Ft. We braced the camera housing to concrete with a short arm to prevent shake.

The problem is at night and on not so perfect visibility days. It is right of the ocean so most mornings around this time a year are foggy.

If you are looking for a high probability of recognition, I would stay away from the long distance LPR.

JH
John Honovich
Mar 24, 2013
IPVM

Sagy, thanks! Can you elaborate on what the image looked like at night and what the key problem(s) were?

Avatar
Sagy Amit
Mar 24, 2013

The main reason we had to mount the camera on the roof was the limitation of running wires. naturally, at night the lighting is limited. the advantage of the height of the camera though is that we are not blinded by headlights.

If you can concentrate a light beam from a flat angle to illuminate the gate area, you would get better results at night.

Avatar
James Talmage
Mar 24, 2013
IPVMU Certified
JH
John Honovich
Mar 24, 2013
IPVM

Interesting find! The tech specs look strong (for low light + long range) though 5kg weight and $4,000+ price may be a barrier for some :)

Avatar
Brian Karas
Mar 24, 2013
Pelican Zero

Interesting find! The tech specs look strong (for low light + long range) though 5kg weight and $4,000+ price may be a barrier for some :)

I was actually going to suggest a 1080p camera, I'm not surprised that Sagy had such good results with one. 500' is not *that* long, but just enough that you're going to start getting into requirements for some less common lenses.

If I were going to attempt this problem I'd think I'd first start with a good 1080p camera with the largest sensor possible. Many of the recent generation units have respectable low-light and WDR specs these days. Then, couple a Canon EOS mount lens to it with the $32 adapter Ethan found in that other thread (I've orderd and received a couple, but haven't had a chance to play with them yet). 150mm isn't really that long in a 35mm format lens, and depending on the sensor you might only need 120mm or so. But, DSLR lens is by nature going to be pretty good quality, and you've got larger elements for better light collection. You can also easily find aftermarket UV/Haze filters, which could be beneficial in this case.

Avatar
James Talmage
Mar 24, 2013
IPVMU Certified

For Brians suggestion, this is probably your best bet. It is cheaper, but by no means cheap ($1,000).

Has anybody demoed those adapters? I missed Ethans post.

Also, be sure to take note of the f-number of the two lenses I've linked to. The Canon (f/2.0) is two stops higher than the Marshall (f/1.0). I would expect the low-light performance of the Marshall lens to be superior.

Can they setup closer than 500 feet for testing? If so use some cheaper lenses to test which f-stop you require before dropping the big bucks.

Avatar
James Talmage
Mar 24, 2013
IPVMU Certified

Also, it doesn't appear either lens is IR corrected. Expect some focus shift between day and night. It may be necessary to have the camera automatically refocus after each shift (so find a camera that allows that).

Avatar
Brian Karas
Mar 24, 2013
Pelican Zero

James - for this application I wouldn't use a a canon branded lens, certainly not an L lens. Sigma makes some really nice EOS lenses about 10% cheaper than similar Canon glass. They're really good qualit, certainly sufficient for an application like this. As you noted, f/stop would be a primary decision factyou for whatever you choose.

Avatar
James Talmage
Mar 24, 2013
IPVMU Certified

Which Sigma lens were you thinking? It seems the f-numbers for those are higher than even the canon for anything that would fit the bill. My biggest concern with higher f-numbers is that you will need to compensate with longer exposures, which is pretty terrible for LPC.

Avatar
Brian Karas
Mar 25, 2013
Pelican Zero

OK, so after looking through some actual options that Canon lens does look like a really good deal :)

For the same price you can get this 50-150mm Sigma but you lose 1 stop.

If you're OK with giving up 1 stop, then you can go to Tokina for about 1/3 the price.

During the day it's likely not an issue. Night time is where you could have problems, and even then, spending the difference on a $600 high-power IR illuminator might be a better option.

Without better details about the parameters of the job it's hard to get much more specific on suggestions.

AI
Andrea Iniguez
Apr 01, 2013

Hi John,

Another option to consider is our SL940 lens with a higher resolution camera. It is IR corrected and is much less expensive than the other options that were being discussed, retailing at $293 MSRP. The SL940 lens at 40mm on a 5mpx camera would provide about 36 pix/ft. That may not be quite enough. You could bump that up to a 10mpx camera and get 48 pixels/foot, though I am not sure of your member's appetite or budget for a 10mpx camera. This lens has been popular for LPR, though I am not sure of the typical distance away it is being used at. We have a couple of example photos on our website with an Avigilon 5mpx camera at 140 to 150ft away, obviously not what you have here. The images demonstrate the very low focus shift at night. The lens is F/1.5, CS mount, auto, or manual iris.

JH
John Honovich
Apr 01, 2013
IPVM

Andrea, if you have to use a higher resolution camera, you increase low light issues, especially since you are suggesting a 5MP or, gulp, 10MP camera. Plus, you now have lower frame rates, especially at 10MP.

A few questions: Is your SL940 lens truly rated for a 10MP camera? What's the f stop of your lens when it is at its most telephoto point, i.e., 40mm?

AI
Andrea Iniguez
Apr 01, 2013

John,

As you know, the resolution question depends on factors such as sensor size, pixel size and %MTF. Our SL940 is designed to resolve a 2.3micron pixel and to have 220 lp/mm in the center consistent with a 5mpx lens. In production we have found we get 300lp/mm or higher in the center consistent with a 10mpx 1.67micron pixel size. We let that fall in half at the edge. Since there aren't many lenses out there that can resolve a pixel of 1.67 micron, we feel the 940 can work - it has been used on 10mpx cameras. The other thing to keep in mind though is field coverage. A common 10mpx sensor size is 1/ 2.3", while the 940 image field is designed for 1/ 2.5" sensor (though it actually is slightly larger), so it doesn't fully cover the smaller sensor without vignetting or dark corners. At 40mm focal length the F # is 1.85. Hope that answers your questions.

JH
John Honovich
Apr 01, 2013
IPVM

Andrea, that does help, especially in terms of the rating/performance of your lens.

The main thing that concerns me is that a 10MP Arecont + an F/1.85 lens is going to have terrible low light performance (primarily due to being 10MP but not helped with the high F stop). Plus, the Arecont 10MP only spec's 7fps and, knowing Arecont, that means more like 4 to 5fps in practice. All this, and you still are only getting 48ppf, which I think is borderline for license plate capture.

However, given how relatively inexpensive it is, it could be worth this integrator trying out.

RG
R Gleason
Jul 09, 2013

What is considered the safe zone for good LPR identification? 40-50 ppf?

JH
John Honovich
Jul 09, 2013
IPVM

Gleason, it depends on whether it the license plates are intending to be read by humans or computers. Humans can deal with less pixels (50ppf is a good minimum for an evenly well illuminated scene). However, if it's automatic, by computers, you would likely need more and that would depend on the system and how accurate you want the recognition to be.

RG
R Gleason
Jul 09, 2013

So for Agent or Aimetis, likely 80 ppf?

JH
John Honovich
Jul 09, 2013
IPVM

Agent who? Agent VI? I was not aware they offer LPR.

As for Aimetis, they have an LPR guide. Here's the relevant section:

I am not sure how tall a license plate character is?

Avatar
Hal Bennick
Jul 10, 2013
Trafficware, a CUBIC Company

I've seen 400' with an analog Pelco running 4CIF 15 about 7 years ago. Crazy long lens, redneck/genius mounting system to reduce shake, and additional lighting for when the sun went down, but it was still cheaper than the trenching that would have been required to move it closer.

So yes, it can be done, but you're going to have to pony up some real $$$ and tinker for a long time to get it right.

New discussion

Ask questions and get answers to your physical security questions from IPVM team members and fellow subscribers.

Newest discussions