proposed license plate camera hack.

[DAP]

Your car windows may have decent IR cut filters built in, in which case best to focus it inside the car with the windows closed. The lens probably won't focus properly in unfiltered daylight.

Looking neat, definitely wants some black on the inside

Painted this morning, needs 24 hours to dry. I went to Home Depot and asked the girl at the paint department for a flat black paint that would stick to aluminum. She suggested a paint that only came in gallon size cans. I pointed out that I only needed to paint about 1 square inch, so she suggested a spray paint. I got Rustolium flat black. Seems to work.

 

[DAP]

I have another idea on how to focus the camera that might be quicker than what I was planning.
The idea is to have the main camera pointed at the side of the rear camera as I turn the lens through peak focus.
Then take footage from both cameras into KDNLive, make sure the audio files of the two videos are synchronized,
Then step through the rear camera video to find the frame with the best photo.
Switch to the front camera video at the same time stamp, and see where the markings on the outside of the lens are relative to a fixed point.
Move the lens back to that point, and use hot melt glue to fix it in place (just enough to keep the lens from moving, so that it is removable in case I decide to change something).

 

[DAP]

My focusing idea worked, but I made the mistake of setting my focus target at the end of my driveway which is only about 25 yards away. I got the camera back together and installed, and went for a test drive, I found the camera had less depth of field than I expected, and was focused much too close.
A side effect of the lack of the IR filter is that all living vegetation comes out pink.
I don't think the field of view is a problem with the 12mm lens, yes it is narrow, but the main camera has a wide field of view and will be quite sufficient.
I will re-focus the camera and try again.

 

[Nigel]

A side effect of the lack of the IR filter is that all living vegetation comes out pink.

Don't set the focus on pink vegetation, that should probably be out of focus if you want number plates to have accurate focus. The pink means that it is seeing IR light, and that probably has a different focus point.

I found the camera had less depth of field than I expected,

The higher the zoom, the less the depth of field, which is what makes 12mm a bit too long, you have to choose what distance to focus on instead of focusing on everything.

 

[Dashmellow]

Don't set the focus on pink vegetation, that should probably be out of focus if you want number plates to have accurate focus. The pink means that it is seeing IR light, and that probably has a different focus point.

That's not quite accurate.

The magenta color cast you see in the image when you remove the IR-cut filter is not infra-red light, per se. It is merely due to the fact that the filter is altering the normal RGB (RED-GREEN-BLUE) color balance that would otherwise be visible. All digital camera sensors employ the RGB additive color model to create images. The combination of RED-GREEN & BLUE light creates the secondary colors of MAGENTA-YELLOW & CYAN. This spectrum of colors combined in the proper amounts creates normal looking colors in images. “White Balance” is achieved when the precise amounts of RED-GREEN & BLUE are mixed together along with the secondary colors to create accurate colors in the image. Digital camera sensors are particularly sensitive to near infra-red and by filtering this part of the spectrum the IR-cut filter allows for the sensor to adequately balance the color model.

Removing the IR-cut filter has the side effect of altering how the MAGENTA portion of the color spectrum manifests. Without the required filter it is throwing the white balance out of whack and this is what causes the “pink” cast in the image. It's not "pink" however. You are seeing the secondary color cast of magenta that the filter would have otherwise removed. "White balance" means that when the additive primary colors of RED-GREEN-BLUE light are present in the proper amounts you get white as seen in the middle of the graphic. You need the IR-cut filter to make that possible on a digital sensor.

The entire image is seeing IR without the cut filter, not just the foliage but the foliage is not reflecting IR light the same was. The IR seen by the camera can manifest in different ways with different colors on a dash cam as seen by the sensor. In some cases your dash cam may even see details not visible with the IR-cut filter.



Foliage is not "out of focus" in an infra-red image. You are witnessing a phenomenon called "The Wood Effect", named after Robert W. Wood, an important American physicist and inventor who pioneered infra-red and ultraviolet photography in the early part of the 20th century.

The Wood effect is when foliage appears bright or white (or magenta) in infrared photographs. It can have a hazy appearance, which can be mistaken for focus. This phenomenon is named after Robert Williams Wood but is often mistaken for "wood" as in trees. What happens is that near infra-red light in the high 600 nanometer wavelength is not absorbed or reflected by chlorophyll. So the light bounces around inside plant cells and either passes through or comes out, including towards the viewer. This often happens with snow, which is why infrared and snow scenes are sometimes mistaken for each other.

What you are seeing is not that the foliage is out of focus in IR but instead is perceived by the camera as literally glowing.

The Wood Effect. (note that it occurs with the red tree foliage as well as the green)

The higher the zoom, the less the depth of field, which is what makes 12mm a bit too long, you have to choose what distance to focus on instead of focusing on everything.

Depends a lot on the particular lens design and aperture in my experience.

Nevertheless, the key to focusing a telephoto dash cam is to spend some time determining the hyper-focal distance the lens is capable of producing rather than picking an object at a specific distance to focus on. Once you get a feel for how your particular lens functions, move the hyper-focal distance forward and backward to optimize the distance between the nearest and farthest points of acceptable focus. In this way available depth of field can be optimized.

As I tried to explain earlier in this thread, the problem I see in using a 12mm lens on this camera is not depth of field but the over-magnification this lens already provides coupled with the fact that it is not well matched with the sensor.

Since the lens is designed for a sensor with a diagonal measurement of 7.19 mm and the sensor in this rear camera has a diagonal measurement of 6.46 mm the result here is that the already high magnification telephoto lens will behave like a much longer focal length telephoto lens than it already is. So, instead of having a 12mm telephoto lens it will have an EFL (effective focal length) perhaps more like a 16mm to 20mm lens. (these are guesses at this point, but you get the idea)

It's not that this lens won't work but the effect of having such a narrow angle-of-view means that whatever you want to capture (such as a license plate) needs the camera to be pointing exactly at the right spot to capture such a tiny window. From my long time experience with telephoto dash cams, you can easily miss the target. You want to capture the road out ahead of you in detail, especially and including cars approaching you in the opposite lane or cars that are not dead center in front of you.

 

[Nigel]

That's not quite accurate.

That is incorrect.

The magenta color cast you see in the image when you remove the IR-cut filter is not infra-red light,

Well, it is true that what you see is not actually IR light, since your eyes can't see IR, and there isn't any IR to see since our monitors don't display IR!
But the sensor is seeing IR and converting it to magenta.

When looking at vegetation, the magenta is the result of the red and blue sensors within the image sensor detecting infrared light from the leaves more brightly than the green sensors detect the green light. If you check the frequency response for the IMX291 sensor, it clearly shows that the red and blue sensors are sensitive to the IR given off by leaves, it is part of its design since it is intended as a security sensor for use in security cameras with IR illumination.

So, "The magenta color cast you see in the image when you remove the IR-cut filter is ", infra-red light converted to magenta light.

What you are seeing is not that the foliage is out of focus in IR but instead is perceived by the camera as literally glowing.

It is very unlikely that this lens will focus infra-red light at the same point as RGB light. This is nothing to do with glowing, just a lack of IR focus correction within the lens. You can buy lenses designed for IR which will focus IR and RGB at the same point, but they tend not to be the cheap ones.

 

[Dashmellow]

Well, it is true that what you see is not actually IR light, since your eyes can't see IR, and there isn't any IR to see since our monitors don't display IR!
But the sensor is seeing IR and converting it to magenta.

The IR -cut filter is only blocking a narrow portion of the spectrum around 650nm while allowing all other visible light to enter. No colors are being "converted" they are merely being seen by the sensor. Without the IR-cut filter you lose the proper color balance corection and the color "shifts" towards the secondary additive color of magenta.

When looking at vegetation, the magenta is the result of the red and blue sensors within the image sensor detecting infrared light from the leaves more brightly than the green sensors detect the green light. If you check the frequency response for the IMX291 sensor, it clearly shows that the red and blue sensors are sensitive to the IR given off by leaves, it is part of its design since it is intended as a security sensor for use in security cameras with IR illumination.

No, that's not how this works either. All digital sensors are sensitive to near IR light. That is the reason that ALL digital cameras have an IR-cut filter.

It is very unlikely that this lens will focus infra-red light at the same point as RGB light. This is nothing to do with glowing, just a lack of IR focus correction within the lens.

Again, no.

You apparently don't how the Wood Effect manifests on camera. The Wood Effect results from the transparency of chlorophyll at wavelengths over 500 nm allowing light to be reflected within the plant cells manifesting in a degree of fluorescence that appears as a "glowing" in photographs.

As for focus correction of IR on a dash cam in a situation like this is a compete non issue. Focusing a dash cam or action camera that has had the IR-cut filter removed is not a problem at all!

IR focus shifts can be a concern under certain circumstances. For example, some CCTV cameras equipped with certain lenses can experience focus shifts at night when under IR illumination during the period when the IR-cut filter is removed from the lens path. If the camera was initially focused under daylight, it may be out of focus at night, or if focused at night it may then appear out of focus in daylight but this is due to the IR illumination used at night which we do not use in front facing dash cams.

When visible light passes through the camera's lens, the focus plane precisely aligns with the image sensor, and so the image is sharp. Infrared light passing through the same lens focuses a bit further from the lens and thus the image renders out of focus on the sensor plane unless the lens is refocused for the infra-red image.

The reason focusing a dash cam that has a lens without an IR-cut filter is not a problem is that the IR-cut filter is only blocking a narrow slice of the electromagnetic spectrum in the near-infra-red range of about 650nm while at the same time allowing visible light to continue to pass through to the sensor. In this case the visible light is what we are focusing on even if the camera is also seeing IR.

IR exclusively shot on film or a modified camera sensor using specialized filter that block ALL visible light but allow only infra-red to pass through. That when focus shift is an issue.

In the “old” days Infra-Red photography was done using black & white or color IR sensitive film and specialized filters that blocked all visible light, allowing only infra-red to pass. A photographer would focus on a scene using visible light before installing the IR filter. This is because once installing this almost pure black filter you could no longer see anything through the viewfinder. After focusing and installing the IR filter the photographer would then manually adjust the lens using IR focus compensations marks that at one time came on almost all camera lenses and occasionally still do. Today, photographers do this digitally but the process is similar and uses black looking IR filters that block all light wave except IR. These are very different than "IR-cut" filters.

Older lenses had Infra Red focus compensation marks that were used AFTER focusing using visible light then installing an IR filter that blocks ALL visible light but passes IR..



I have been experimenting with IR lenses on dash cams for more than 5 years now and I can tell you with certainty that focusing without an IR filter is not a problem.

Anyone who is in interested in this topic and wants to learn more about it or simply wants to see many examples of an Infra-Red enabled dash cam using a lens without an IR-cut filter (and with too) should visit the following thread from 2017. Among many other things you'll see the first telephoto dash cam images ever posted to DashCamTalk.

Mobius Varifocal Zoom IR​

You'll see many tack sharp dash cam images where I merely focused the camera using the analogue AV output from the Mobius camera connected to a 7 inch 720p portable monitor via its RCA AV inputs. Works perfectly and easily.





I achieved spectacularly sharp images with zero focus shifts. For those who don't feel like visiting the thread, here are a few examples.

This whole notion of IR focus shifts or difficulties with focusing a dash cam that has no IR-cut filter is complete nonsense!

Hands on experience will teach you far more than what you think you may know from reading things on the internet.








Here's one after installing an IR-cut filter.

​

 

[Dashmellow]

BTW, one more thing to add here.

If IR focus shifts were really an issue when not using an IR-cut filter on a dash camera in daylight or at night with IR illumination, Viofo would not have been able to achieve such excellent results with the T-130 interior camera!

Here are some examples, courtesy of @Vortex Radar that popped up on a search or "Viofo T-130 interior views".





Hmmm......visible light capture as well as IR.

 

[DAP]

Daytime test footage:

I got the focus much better here.
The video shows 5 minutes of daytime video with the rear camera followed by the same 5 minutes with the main camera showing the same scene so a direct comparison can be made.
License plates can be read at further distances, but not by an impressive amount. Most of that is due to the rear camera sensor being only 1080p

As a side note, the car triggered a false positive forward collision warning shortly after this video that cost me 3 points on my safety score.
Youtube is estimating another hour to get this uploaded.

 

[Nigel]

If IR focus shifts were really an issue when not using an IR-cut filter on a dash camera in daylight or at night with IR illumination, Viofo would not have been able to achieve such excellent results with the T-130 interior camera!

Unlike you, they understand how the optics and sensors work!

Yes, Viofo have achieved excellent results for the IR on the T130, best IR dashcam image I have ever seen, because on the T130 they used an Omnivision image sensor specifically designed for IR, featuring "Omnivision Nyxel infrared technology", along with a lens specifically designed to focus IR light correctly at IR wavelengths, something standard lenses do not do.

The Wood Effect results from the transparency of chlorophyll at wavelengths over 500 nm allowing light to be reflected within the plant cells manifesting in a degree of fluorescence that appears as a "glowing" in photographs.

Where on earth did you read that the Wood effect has anything to do with fluorescence?

 

[Dashmellow]

Unlike you, they understand how the optics and sensors work!

Yes, Viofo have achieved excellent results for the IR on the T130, best IR dashcam image I have ever seen, because on the T130 they used an Omnivision image sensor specifically designed for IR, featuring "Omnivision Nyxel infrared technology", along with a lens specifically designed to focus IR light correctly at IR wavelengths, something standard lenses do not do.

I understand a lot more than you how this works because unlike you I've had years of hands-on experience with these things.

I also see you've ignored my comments on focusing a camera without an IR filter, which debunk your claims about focus shifts so, of course you change the subject and turn to adolescent insults.

Where on earth did you read that the Wood effect has anything to do with fluorescence?

From the erroneous remarks you've made in this thread, it is clear you never even heard of the Wood Effect until I posted about it. I've already explained about how chlorophyll and foliage react under IR light, so I won't bother with your question. I will add however that the effect from chlorophyll fluorescence due to light scattering in every direction inside of leaves is subtle in IR photography and it is not the cause of the brightness you see. It does however create the glow you claim is an IR focus issue.

 

[TonyM]

As a side note, the car triggered a false positive forward collision warning shortly after this video that cost me 3 points on my safety score.

I know it's off-topic for this thread, but can you explain what this means? I'm not familiar with a safety score, nor how your car could trigger a collision warning, either true or false. Maybe I'm just too old or behind the times to understand.

 

[DAP]

I know it's off-topic for this thread, but can you explain what this means? I'm not familiar with a safety score, nor how your car could trigger a collision warning, either true or false. Maybe I'm just too old or behind the times to understand.

Safety score is something Tesla insurance uses to set the price of their insurance in states that allow it. It is also used to determine who get into the full self drive beta test. I am attempting to get into the beta test, so I have to keep my safety score as high as I can.
The way safety score works, if you do a small amount of driving in a particular day, and get a forward collision warning, it has a HUGE affect on your safety score. Because of that forward collision false alarm today, I just went on a 40 mile drive and gained back two of the points I lost from the forward collision warning.
Here is Tesla's page on the safety score: https://www.tesla.com/support/safety-score
A big problem with it is false positives.
Another big problem with it is that California times their stop lights in such a way that if you get a yellow at just the wrong time, you have to choose between a hard braking penalty and running the red light because the light will be red by the time you get there, and stopping requires exceeding the G force limits in the safety score.
Following distance is also nasty. If you drive 100 miles with nobody in front of you, then someone merges in front of you, you get a large following penalty. This is because the following penalty is calculated based on the ratio of good following (2-3 seconds following distance) to bad following (less than 2 seconds).
Your safety score is very much dependent on how other people drive around you, and random false positives.
It is really frustrating trying to keep it high, and kills the fun of driving.

 

[Dashmellow]

It is really frustrating trying to keep it high, and kills the fun of driving.

There is an entrance ramp to a major interstate highway near where I live where two lanes coming from opposite directions merge into one as they head up the ramp to the highway heading north. One lane has a yield sign while the other one always has the right-of-way.

Recently, I was heading up the ramp in the lane that has the right-of-way when suddenly a Tesla came flying right through the yield sign without even slowing down and cut me off on his way up the ramp. It was a tourist from out of state which is often typical behavior we see here all the time. Ordinarily, I would have been really pissed off but I had so much fun witnessing just how fast this guy was able to accelerate up the ramp in front of me when he cut me off that I actually enjoyed the experience. Obviously, this guy wasn't going for any kind of safety score. There is a 16 bay Tesla charging station right nearby where many out of staters stop for a charge and then get back on the highway and I gather this guy had just charged up.

 

[DAP]

I have added a night time test.

Interesting things to note:
Vegetation shows up green when illuminated by my LED headlights.
Incandescent light sources appear pinkish white regardless of the color filter in front of them (headlights & tail lights both look the same color on cars with incandescent bulbs).
My eyes can see detail where both cameras see only black.
My eyes can see detail in license plates where the cameras only see white.
There is significant motion blur in moving objects, unlike the daytime capture.

I'm going to change the exposure on the rear camera to -2. All of the license plates are over exposed at night. The sole purpose of this 12mm lens camera is to capture license plates, so under exposing the rest of the image will not be a problem since the scene is also captured by the main camera.

Youtube is estimating 35 minutes for the upload to complete.

 

[Harsh]

You need a better lens. Try the 5MP ones from Fulekan or Treeye.

 

[DAP]

You need a better lens. Try the 5MP ones from Fulekan or Treeye.

This lens claims to be a 5MP lens. The sensor is only 1080P though. The videos are 4K, because the second half of the video is the main 4K camera. The first half is scaled up to match.

 

[TonyM]

Interesting things to note:
Vegetation shows up green when illuminated by my LED headlights.
Incandescent light sources appear pinkish white regardless of the color filter in front of them (headlights & tail lights both look the same color on cars with incandescent bulbs).
My eyes can see detail where both cameras see only black.
My eyes can see detail in license plates where the cameras only see white.
There is significant motion blur in moving objects, unlike the daytime capture.

I'm going to change the exposure on the rear camera to -2. All of the license plates are over exposed at night. The sole purpose of this 12mm lens camera is to capture license plates, so under exposing the rest of the image will not be a problem since the scene is also captured by the main camera.

Whilst the IMX291 is good for low light capture, it is not the best sensor for dynamic range or fine detail, at least not in the way that Viofo have setup their firmware with high contrast and sharpening. Unfortunately there's not much you can do about the A129 Pro rear camera. I did read that recent FW renders shadows darker than earlier versions, so you could try going back to V2.4 or V2.5. I don't think BCHobbyist made any modified FW for the A129 Pro.

Motion blur will be exacerbated by the focal length - objects move across the sensor much faster with a 12mm lens than a 3mm lens, and any slewing or vibration from your car will be amplified by the increased magnification. That's another reason why some of us have settled on telephoto focal lengths around 6-8mm.

Dropping the exposure for number plate capture at night will only get you so far, because it will affect the daytime capture in anything less than bright sunny conditions. And you will get even more very dark or black areas in the video.

Whilst the video footage without an IR-cut filter is interesting, as a driving recorder I think it is useful to have somewhat accurate colours, so your camera can differentiate between tail lights and brake lights, or show the difference between black, blue and grey cars. If you really are only interested in licence plates then colour accuracy is less important.

 

[DAP]

Whilst the IMX291 is good for low light capture, it is not the best sensor for dynamic range or fine detail, at least not in the way that Viofo have setup their firmware with high contrast and sharpening. Unfortunately there's not much you can do about the A129 Pro rear camera. I did read that recent FW renders shadows darker than earlier versions, so you could try going back to V2.4 or V2.5. I don't think BCHobbyist made any modified FW for the A129 Pro.

Motion blur will be exacerbated by the focal length - objects move across the sensor much faster with a 12mm lens than a 3mm lens, and any slewing or vibration from your car will be amplified by the increased magnification. That's another reason why some of us have settled on telephoto focal lengths around 6-8mm.

Dropping the exposure for number plate capture at night will only get you so far, because it will affect the daytime capture in anything less than bright sunny conditions. And you will get even more very dark or black areas in the video.

Whilst the video footage without an IR-cut filter is interesting, as a driving recorder I think it is useful to have somewhat accurate colours, so your camera can differentiate between tail lights and brake lights, or show the difference between black, blue and grey cars. If you really are only interested in licence plates then colour accuracy is less important.

Since there are two cameras aimed forward, one being the un-modified A129 pro 4K camera, the telephoto's ONLY job is capturing license plates. So color accuracy and shadow detail are irrelevant.
Dropping the exposure has two advantages, it will prevent license plates from being over exposed at night, and shorten the exposure time at night, reducing motion blur. This might cause problems during the day, I have not tested this yet.
The main camera is pretty much useless at night for capturing license plates, does OK during the day.
The drawbacks of this setup are that the rear camera is only 1080p, the camera runs hotter due to having to compress two video streams simultaneously, and it reduces the amount of video time that can be stored on the memory card.

 

[DAP]

Daytime test footage with exposure set to -2. Seems somewhat detrimental to daytime footage.

Youtube is estimating that the upload will complete in about an hour.