Poor IR contrast in new Tennessee license plates

[wtimothyholman]

@EMPIRETECANDY - any word on engineering? I can only imagine what kind of market penetration you could get if you could market a camera that is sensitive to 730nm, along with 730nm IR LEDs. There appears to be a real lack of such consumer level cameras - and as more states move to these printed, white-lettered plates, the demand will only increase. Early bird gets the worm!

I second this. I've had people in law enforcement ask me if I know of any 730nm or 740nm cameras that can be purchased for license plate capture. The situation is already bad in Tennessee, and it will only get worse.

Dahua could make a lot of money if they offered a -Z12E varifocal camera with 730nm or 740nm IR. I would buy two of them immediately, even if they cost $500 each.

 

[tech_junkie]

I second this. I've had people in law enforcement ask me if I know of any 730nm or 740nm cameras that can be purchased for license plate capture. The situation is already bad in Tennessee, and it will only get worse.

Dahua could make a lot of money if they offered a -Z12E varifocal camera with 730nm or 740nm IR. I would buy two of them immediately, even if they cost $500 each.

The possible problem might be it needing a 1.2-1.3 aperture instead of shorter wavelength. Larger aperture = more optical energy reflected in regardless of wavelength.

Near red grow lights are that wavelengths. But anything under 780nm is a visible color.

 

[erkme73]

The possible problem might be it needing a 1.2-1.3 aperture instead of shorter wavelength. Larger aperture = more optical energy reflected in regardless of wavelength.

Near red grow lights are that wavelengths. But anything under 780nm is a visible color.

I'm ok with a dim red/visible led if it means I can read the tags. Whether it's a change of lens aperture or a different sensor (or both), we need to get someone to develop a solution asap. Whoever gets this to market first will likely make a killing.

 

[wtimothyholman]

I'm ok with a dim red/visible led if it means I can read the tags. Whether it's a change of lens aperture or a different sensor (or both), we need to get someone to develop a solution asap. Whoever gets this to market first will likely make a killing.

My only concern is that it not distract passing drivers. In that respect a pulsed white light may be better in a residential neighborhood, but I’ll take anything that works.

 

[sebastiantombs]

I see getting accurate triggering as a problem. An example being one car following another fairly closely or, alternately, two cars from opposite directions at the same time or in a short time. Getting a visible light to pulse at just the right time in either situation is going to be a very daunting task. Remember toll booths have a distinct advantage because the direction of travel is one way only.

 

[wtimothyholman]

I see getting accurate triggering as a problem. An example being one car following another fairly closely or, alternately, two cars from opposite directions at the same time or in a short time. Getting a visible light to pulse at just the right time in either situation is going to be a very daunting task. Remember toll booths have a distinct advantage because the direction of travel is one way only.

No, the light would run continuously. But if it has a low duty cycle and a high frequency, it will simply look like a regular white light to anybody who sees it.

 

[erkme73]

I see getting accurate triggering as a problem. An example being one car following another fairly closely or, alternately, two cars from opposite directions at the same time or in a short time. Getting a visible light to pulse at just the right time in either situation is going to be a very daunting task. Remember toll booths have a distinct advantage because the direction of travel is one way only.

I wonder about this too. When I go back to review a "beautiful, full-sky lightning bolt", I find that the camera doesn't capture it - or if it does, it's half a frame, over exposed, and not really usable. Perhaps the pulsing/flashing of the LED could be timed with the shutter speed to where they fire at the same rate, but if there's any variability, some frames will be dark, while others will be illuminated. And we know how murphy woks... the dark frames will be just as the bank robber drives by....

 

[wtimothyholman]

I wonder about this too. When I go back to review a "beautiful, full-sky lightning bolt", I find that the camera doesn't capture it - or if it does, it's half a frame, over exposed, and not really usable. Perhaps the pulsing/flashing of the LED could be timed with the shutter speed to where they fire at the same rate, but if there's any variability, some frames will be dark, while others will be illuminated. And we know how murphy woks... the dark frames will be just as the bank robber drives by....

The ideal case would be where the LED is integrated with the camera to ensure that the shutter speed and frame rate are coordinated with the pulses of light. I typically get multiple frames of the same license plate as it moves across my field of view. Assuming you select a pulse frequency that is somewhat different than the frame rate, you will probably get at least one or two good images.

 

[wittaj]

When I pulled up the specs on the unit, it says it has the capability to sync it with the shutter, so is the z12e capable of communicating with it?

 

[wtimothyholman]

There are times when you kick yourself and say, "Why didn't I think of this before?"

I am happy to report success with my initial test of the Dahua -Z12E camera in color mode. The shutter speed was set to 1/2000 second, but the camera was forced into color mode, with IR turned off. In effect, the camera was in daytime mode. I then took a small LED flashlight and held it underneath the camera, trying to keep the beam as parallel to the direction of the camera as possible.

Below is the result, using my own plate as my wife drove around the block. Unfortunately the beam width was fairly wide and the vehicle was about 70 feet away, so I didn't get a lot of illumination on the plate, but it was sufficient for OpenALPR to read it. And just to be certain, I turned off the flashlight and verified that the plate could no longer be read on the next pass.

So white LED illumination does look like a possible solution. The trick will be to implement something like the Axton Pulsar illuminator, where very bright pulses of light are generated for very brief durations, such that the average brightness is not distracting to passing drivers. Ideally the LEDs could be integrated into the camera, with 1 to 2 ms light pulses synchronized to the camera's frame rate.

This is a modification that Dahua engineers could incorporate into a version of the -Z12E camera without having to change the optics. (@EMPIRETECANDY, please take note of this!) There would also be the advantage of eliminating the need for the day / night refocusing transition of the LPR camera, beyond turning on and off the white LEDs at sunset and sunrise.

My next step is to try to figure out some sort of low-cost alternative to the Axton Pulsar. First, I'm going to purchase an LED flashlight that can be focused to a narrow beam and do some tests with it. Second, I'm going to look into incorporating some type of pulse width modulation for the flashlight.

 

[wtimothyholman]

I'm looking at modifying some very low-cost LED strobe lights as illuminators. Assuming that the camera is free running at 60 fps with a shutter time of 1/2000 seconds = 0.5 ms, and that the strobe light flashes 100 times a second with an "on" time of 2 ms (20% duty cycle), I ran a quick simulation of the intersection of those intervals. The result is shown below.

Without any synchronization between the camera and strobe light, I'll effectively get about 20 fps of illuminated plates (illustrated by the pulses rising to a value of 1 V). The trick is to ensure that the camera frame rate and the strobe frequency are significantly different, and one is not a harmonic of the other.



The optimal situation would be where the camera itself triggers the strobe just before the shutter opens, but I'm not sure I can do that with my -Z12E. Regardless, I'll still get a sufficient intersection between the two frequencies to further test this idea.

 

[wittaj]

If you can figure out how to run it in color without it looking like a stadium at night or causing epileptic seizures due to the strobing, this would be a game changer for sure. More temp paper plates would be picked up, as would dirty tags, and just as important, could provide for an opportunity to run other cameras in color with a slower shutter instead of infrared- but I would guess some trial and error would need to be done to not have shutter and pulse out of whack and creating a useless image.

 

[wtimothyholman]

If you can figure out how to run it in color without it looking like a stadium at night or causing epileptic seizures due to the strobing, this would be a game changer for sure. More temp paper plates would be picked up, as would dirty tags, and just as important, could provide for an opportunity to run other cameras in color with a slower shutter instead of infrared- but I would guess some trial and error would need to be done to not have shutter and pulse out of whack and creating a useless image.

From my simulation, an fps of 60 would intersect with a 100 Hz strobe between 19 and 21 times per second, no matter the phase shift of either. What you don't want is to run both at 60, or one at 30 and the other at 60. Then you have the possibility of extended "dark" times when the shutter and strobe don't intersect.

Also, a 100 Hz strobe can only be perceived by a tiny fraction of the population. Most can't see it. Fluorescent lights have an effective strobe frequency of either 60 Hz or 120 Hz (depending on the lighting), so if driving past a storefront with a fluorescent display doesn't trigger an epileptic seizure, this won't either. One could also increase the strobe frequency to 200 Hz with a pulse duration of 1 ms, and even push it past that point.

As to how bright the light must be: it will certainly be noticeable, but it will be white, not red. The lower duty cycle will make it look much dimmer to the eye (on average) than another light running at continuous full intensity. That's what I need to experiment with. You don't want the white LED light to be bright enough to distract a passing driver.

 

[wittaj]

As long as it doesn't look like a disco light LOL or like a misaligned floodlight blinding the driver, I think it would be ok?

Now I am wondering how close I can "misalign" my floodlight and get a color capture? I know what I will be trying tonight lol

 

[biggen]

I've driven through plenty of toll stations at night that use pulsing white light for images of cars/plates and it's very noticeable. I'd think the Karens in the neighborhood would raise Cain about using something like that. I'd never get my wife onboard that is for sure. Still interested in your plan however.

 

[wtimothyholman]

I've driven through plenty of toll stations at night that use pulsing white light for images of cars/plates and it's very noticeable. I'd think the Karens in the neighborhood would raise Cain about using something like that. I'd never get my wife onboard that is for sure. Still interested in your plan however.

Not if it's continuously pulsing at 50 to 60 Hz, or higher. It will be no different than a fluorescent light. The light will not turn "on" or "off" to the human eye; it will just be "on" constantly.

I've seen enough from my tests so far to think that it is at least feasible.

 

[biggen]

Not if it's continuously pulsing at 50 to 60 Hz, or higher. It will be no different than a fluorescent light. The light will not turn "on" or "off" to the human eye; it will just be "on" constantly.

I've seen enough from my tests so far to think that it is at least feasible.

The ones they use in toll stations must act like camera flashes then because I can see them "flash" when a car goes through the toll lane.

Still, the constantly on may be an issue for some unfortunately.

 

[wittaj]

I would think the ones at toll booths are flashes.

Too bad the cycling of the light makes it look like constant "on" instead of "off" lol.

I do have instances where the light cycle of someone's headlights appear like they are off, so it would be important to figure out how to prevent that with the camera and the pulsed light.

I guess the question will come down to how bright it is. If it would look the same as a 3000 lumen floodlight then why not just point a $25 night chaser floodlight that direction?

 

[wtimothyholman]

I would think the ones at toll booths are flashes.

Too bad the cycling of the light makes it look like constant "on" instead of "off" lol.

I do have instances where the light cycle of someone's headlights appear like they are off, so it would be important to figure out how to prevent that with the camera and the pulsed light.

I guess the question will come down to how bright it is. If it would look the same as a 3000 lumen floodlight then why not just point a $25 night chaser floodlight that direction?

The idea is that it would be as effective as a 3000 lumen floodlight to the camera, but only look like a 300 lumen light to the human eye. These lights will be aimed at the road for LPR, and the last thing you want is people complaining that your floodlight is blinding them as they drive by your home.

White lights are nothing unusual in a residential area (e.g. landscape lights, porch lights, etc.). As long as it's not too bright, a pulsed LPR light would be noticed but ignored.

 

[wtimothyholman]

When I pulled up the specs on the unit, it says it has the capability to sync it with the shutter, so is the z12e capable of communicating with it?

That's a good question, and one I can't answer. While Axton says that their Pulsar unit supports some camera brands, I've found no indication that Dahua is among them.