How important is license plate capture?

I wonder if they could make a dual camera, where both are front facing but the one camera a plate camera.
So you had 1 camera to give you the perspective of what is going on, and another camera synced to that but tailored to only do plate capture.
I do find that with a long range "Zoom" modified dashcam, you have a longer time / more frames in which you have a chance of a plate capture, and it is further out front so the oncoming car / plate are still at that angle coming more towards you than zipping right past you off to the side where a regular dashcam have its sweet spot pretty close up., so i feel motion blur are less of a issue for the zoom camera..
 
I wonder if they could make a dual camera, where both are front facing but the one camera a plate camera.
So you had 1 camera to give you the perspective of what is going on, and another camera synced to that but tailored to only do plate capture.
I do find that with a long range "Zoom" modified dashcam, you have a longer time / more frames in which you have a chance of a plate capture, and it is further out front so the oncoming car / plate are still at that angle coming more towards you than zipping right past you off to the side where a regular dashcam have its sweet spot pretty close up., so i feel motion blur are less of a issue for the zoom camera..
We discussed this before, and I think concluded that it is better to have one 4K sensor than two 1080 sensors where one has a narrow angle number plate reading lens. Mainly because with the narrow angle lens the number plate often falls off the side of the image.
 
Too many "ifs" :( If the FOV isn't wide enough... if the main cam has better resolution...and more ;)

I wish I were in a position to 'play' with this, as there's still a lot of uncovered ground with the telephoto/plate capture testing as yet undone. Tough enough just finding a tele lens and cam combo which works well. Perhaps it should be reversed, with the main cam 1080P (which is plenty good for anything else) and the 4K for telephoto/plates :cool:

And then there's the rear of the vehicle to consider too, in case the front cam missed an oncoming plate or as it is here cars have rear plates only :whistle: Never the "perfect" cam but we're getting closer!

Phil
 
I'd rather have poor quality general video but visible licence plates. Even the lowest end over-compressed 480p is enough to determine blame in an accident. Tune the camera to capture plates as the priority.

I was hoping to find a camera that was open enough to write my own software for, or at least modify the existing software. Like some DSLRs can run modified firmware that gives you loads of extra features.

I might get a couple of cameras off AliExpress and try with a Raspberry Pi. They have lots of them that claim wide angle and good low light performance, although they are probably just the same as in most dashcams.
 
I'm wondering if something like these would work:


6 or 12 MP using a nice Sony large format sensor. You need to supply your own fisheye lens. Also probably some kind of image stabilisation mount.

The first one is a 1/1.8" sensor and the second one a 1/1.7" sensor. Dashcams are generally 1/3.5" or something like that. Capturing say 15 fps at 12 bit I think there is a decent chance you will be able to read licence plates with it.

I wish someone would make a dashcam with a sensor like that. Why do even £500 cameras have diddy little sensors?
 
If he leave the scene YES, but often with crashes at least to my knowledge the cars just sit there and aint going nowhere.
You can easy determine who is to blame with a 2-3-4 car crash captured on video, even if none of the cars have a plate on them.
IMO Americans must have dual systems as just one plate on the car seem to be a norm over there, so you need that rear camera badly. more easy for us EU boys with 2 large readable plates with no silly graffiix ASO on them.
 
There are times where someone is to blame even if they are not directly involved in the crash. Like say they pull out and you brake hard to avoid hitting them. You skid and hit a parked car as a result. At least in the UK they are responsible because although they didn't hit you they caused you to crash.

Without licence plate capture they can easily just pretend not to be involved and drive off.
 
Indeed, it is also why we all should be responsible and vigilant witnesses, and not just drive past and think " poor sob he did not have a chance"
Making sure the right part get to pay up are one of the main reasons i got into dashcams, not least in regard to the overprotected and shameless and bad cyclists here.
CUZ touch one of them with your car and you probably have to retake your driving license,,,, and i do not want to do that if if was his / her fault i touched.
 
Larger sensors which can do better with low-light are probably coming to dashcams, but it's anyone's guess when we will get them.
We already have larger sensors,
  • The IMX322 that was popular a couple of years ago is 16.26 mm².
  • The IMX292 that is currently popular is 17.44 mm².
  • The IMX317 in the Viofo A129 Pro is 21.77 mm².
  • The IMX334 in the Thinkware U1000 is 33.18 mm².
There is more to it than just increasing size though, the Viofo A129 Pro appears to be beating the Thinkware U1000 on image quality and plate readability.

Increasing size also comes with increasing cost, sensor cost tends to be proportional to pixel area, and lens cost tends to be proportional to pixel area², which partly explains why the U1000 costs twice the A129 Pro.

Next step is for someone to make a dashcam using the IMX485, which has a pixel area of 69.76 mm², and according to Sony a sensitivity of 3.3x the Thinkware U1000's IMX334. Should make a nice dashcam, but it's not going to be cheap!

Alternatively we can use improved versions of existing size sensors, but the improvements in sensitivity are getting very small, we seem to be close to the limit for existing sensor technology. We need completely new technology capable of turning starlight into daylight, real Starvis! :




IR is certainly possible with external illuminators, but that layer of complexity is too far for most people. And there's always stolen or altered plates to contend with. As individual experimenters this is very interesting, but it will be nigh on forever before it happens out-of-the-box.
Using IR for external illumination requires slightly higher light levels than for normal illumination since the sensors are a little less sensitive to IR, so if you have normal 55W headlamps in your car, you will want 65W IR headlamps alongside them to get a similarly sharp image. That level of IR light is very dangerous to human eyes so can't be used, and if used for parking mode would empty a car battery in 30 minutes so is also impractical. IR security cameras don't need that amount because they are static so can use very low shutter speeds, and maybe only do 3fps anyway so they are not limited to 1/30th second like a 30fps dashcam is.
 
We shouldn't need night vision or IR illumination for licence plates. In Europe they have to be lit, it's mandatory. Is it not the same elsewhere?

That's actually the problem. Being lit they tend to just get washed out at night, saturating the camera sensor. Maybe manufacturers could implement a special "plate mode" that tunes the video for licence plate capture at the expense of dynamic range. I don't care so much if the video is a bit dark or there isn't much shadow detail if I can read the plate.
 
Oh and looking at review footage of the Viofo A129 Pro and the Thinkware U1000 they both look worse in low light than my old Seezeus. No chance of reading plates, loads of noise on those 4k sensors.
 
The IMX 485 etc is the "larger" I'm referring to. The physical size of a sensor means a limit to the light it can capture, so a larger sensor can capture more light and with good firmware that should allow for more ability to 'fine tune' the image to see more of what we want and less of what we don't want. On costs, all new technology (including new applications of the existing tech) starts at higher costs. As manufacturing improves and the market increases the cost comes down.

If costs and size were no object we already have military-grade systems which can "see" at night almost as well as in daytime, but we're not going to get anything like that in a simple dashcam. We're doing pretty darn good right now for what most people really need, which is first a means of proving that we have no part in fault, then secondarily a means to identify other vehicles. It's really car insurance that does the 'protecting'; our cams only enhance that process. So even as nice as positive plate capture would be, IMHO it's not as necessary as it would seem to be.

Phil
 
The IMX 485 etc is the "larger" I'm referring to. The physical size of a sensor means a limit to the light it can capture, so a larger sensor can capture more light
Actually it is not that simple, there is a limit to how big we can have the sensors, as you increase the sensor size you need to increase the focal length of the lens to maintain the field of view, and increasing the focal length of the lens decreases the depth of field, at some point you have to start decreasing the lens aperture to maintain depth of field, otherwise you would have to start focusing on individual number plates, which isn't possible in a dashcam, everything has to be in focus. Decreasing aperture puts a limit on the amount of light you can collect, and because of that, once you reach the limit, the bigger sensors are not actually going to collect more light.

In recent years the aperture of our dashcam lenses has been decreasing, our action cameras are about F2.8, five years ago dashcams were F2.2, then we moved to F1.8, now F1.6 is popular, and as lens technology improves I think we will soon be seeing F1.4 as normal for the top dashcams. The result of this decrease is that the maximum useful sensor size has been decreasing. The IMX 485 is probably just under the current limit, so it is the ideal sensor size for a dashcam, unless someone comes up with a F1.0 dashcam, in which case it may be too big!
 
Lets be honest, the odds of you being in an accident where clearly seeing the plates and need it is very, very rare. The main purpose is to see people, objects and other autos and to absolve you of fault in the case of an accident. Parking lot incidences where you are not present rarely result in you even seeing the car that hits yous plate number. As much as I would like it as well, unless you go with a narrower FOV, you just might not be able to resolve the plate number.
 
Lets be honest, the odds of you being in an accident where clearly seeing the plates and need it is very, very rare. The main purpose is to see people, objects and other autos and to absolve you of fault in the case of an accident. Parking lot incidences where you are not present rarely result in you even seeing the car that hits yous plate number. As much as I would like it as well, unless you go with a narrower FOV, you just might not be able to resolve the plate number.

Car parks and "if not for" incidents. I don't know if you have a rule like that in the US but in the UK if you say swerve to avoid someone and then crash they are at fault even if they didn't hit you. They might just drive off so it's important to get their details. Of course often the camera won't even have a good view of them, e.g. if they start pulling out at a right angle so the plate is at nearly 90 degree to you.
 
A few days ago I learned that the plate number is very important key evidence that without nobdy is interested in it. However, to pay $400 for the dash cam doesn't make any sence to me if your deductible is $250-500 and you are not at fault. Also, there is no guarantee if it will record the footage suitable to seek reimbursement for the damages caused to the property. So, reasonably priced cams or set ups are good enough to meet the needs.

For the argument sake the cases like mine are rare. Who would leave overnight a luxury vehicle in the streets of a busy city where the amount of drivers who are on something being high is staggering. Mostly people here own the clunkers and would care less if more or less damage appears on the car.
 
Back
Top