How important is license plate capture?

[SawMaster]

Then if you think phones are the better solution you should use them instead of coming here to a dashcam forum to essentially bash the products we use

Phil

 

[comadose]

A phone with dashcam software might not be a bad solution actually... Depends how fast it boots and if you can get it to power on automatically with the ignition.

 

[DT MI]

A phone with dashcam software might not be a bad solution actually... Depends how fast it boots and if you can get it to power on automatically with the ignition.

There's an entire sub-forum on that very subject - https://dashcamtalk.com/forum/forums/smartphones-as-dash-cams.20/

 

[Nigel]

A phone with dashcam software might not be a bad solution actually... Depends how fast it boots and if you can get it to power on automatically with the ignition.

Looking forward to seeing your fantastic dashphone video, or maybe laughing at the poor result...

 

[andrew124C41]

Hi everyone,

I am new here, and just started the process of researching my dashcam purchase. From my brief research, it seems that most dashcams currently on the market are not capable of reliably capturing US license plates when traveling at highway speeds, even during the day.

1080p simply isn't enough resolution when we're talking about a 140°+ field of view to render a readable license plate, unless it's very close and straight ahead. The 4k DR900S is better, but its resolving power is compromised by its extremely wide 162° FOV, so it's not measurably better than the 1440p Viofo cameras. Additionally, the bitrate on most dashcam models is far too low, and still frames turn into mush. I do understand that dashcams have limited processing power, so compression settings have to be tweaked for speed over quality, but this problem can be fixed by increasing the bitrate. While 100Mbps will drastically cut recording time, it will greatly increase the odds of identifying a license plate in the footage. This is not for everyone, but it seems that it should at least be an option, instead of the 25Mbps that the high end cameras top out at.

Finally, in reviews, few people discuss dynamic range of sensors, and I can't find anyone who has actually measured them at all. Most people talk about the "brightness" of footage, and use this as a metric to compare different cameras, but brighter footage does not mean that the quality is better. The only metrics that matter for forensics are resolution and dynamic range (signal-to-noise ratio). Furthermore, darker footage is often better than brighter footage, given the reflective nature of license plates. Once an area is blown out, no details can be recovered.

I have a two-part question:

1. Is it actually important in most cases to be able to capture a license plate? It matters for hit-and-runs, but how common are those events compared to "normal" incidents when dashcam footage is used help arbitrate a he said, she said situation?
2. If it actually is important, what is the best resolving dashcam currently on the market for both day and night recording, after tweaking with custom firmware?

Hi everyone,

I am new here, and just started the process of researching my dashcam purchase. From my brief research, it seems that most dashcams currently on the market are not capable of reliably capturing US license plates when traveling at highway speeds, even during the day.

1080p simply isn't enough resolution when we're talking about a 140°+ field of view to render a readable license plate, unless it's very close and straight ahead. The 4k DR900S is better, but its resolving power is compromised by its extremely wide 162° FOV, so it's not measurably better than the 1440p Viofo cameras. Additionally, the bitrate on most dashcam models is far too low, and still frames turn into mush. I do understand that dashcams have limited processing power, so compression settings have to be tweaked for speed over quality, but this problem can be fixed by increasing the bitrate. While 100Mbps will drastically cut recording time, it will greatly increase the odds of identifying a license plate in the footage. This is not for everyone, but it seems that it should at least be an option, instead of the 25Mbps that the high end cameras top out at.

Finally, in reviews, few people discuss dynamic range of sensors, and I can't find anyone who has actually measured them at all. Most people talk about the "brightness" of footage, and use this as a metric to compare different cameras, but brighter footage does not mean that the quality is better. The only metrics that matter for forensics are resolution and dynamic range (signal-to-noise ratio). Furthermore, darker footage is often better than brighter footage, given the reflective nature of license plates. Once an area is blown out, no details can be recovered.

I have a two-part question:

1. Is it actually important in most cases to be able to capture a license plate? It matters for hit-and-runs, but how common are those events compared to "normal" incidents when dashcam footage is used help arbitrate a he said, she said situation?
2. If it actually is important, what is the best resolving dashcam currently on the market for both day and night recording, after tweaking with custom firmware?

My wife and I just went through this with a Blackvue 590WD. You would think for the price it could capture a plate. We were driving at night and a guy passed us on the right shoulder at high speed. We wanted to submit the plate number to the police but we got nothing. In addition, while parking in NYC, a guy parked too close. There was no damage, but we realized that we could not see the plate had there been a problem. We just purchased the camera and I am going to return it to Blackboxmy car. The dashcam is absolutely, totally worthless.

I suspect that there are none on the market that will work because a frame rate of even 60 might not catch it even at 4 K.

 

[relive]

My wife and I just went through this with a Blackvue 590WD. You would think for the price it could capture a plate. We were driving at night and a guy passed us on the right shoulder at high speed. We wanted to submit the plate number to the police but we got nothing. In addition, while parking in NYC, a guy parked too close. There was no damage, but we realized that we could not see the plate had there been a problem. We just purchased the camera and I am going to return it to Blackboxmy car. The dashcam is absolutely, totally worthless.

I suspect that there are none on the market that will work because a frame rate of even 60 might not catch it even at 4 K.

1. Blackvue cameras are not known for their image quality. People who buy them because they like the design, features and brand value. There are multiple threads on this forum complaining about image quality of their flagship DR900S.

2. 1080p simply doesn't have the resolution to capture plates from more than ~25 feet away with a wide angle lens. I have a Viofo A129 Pro now, and even it has problems reading the plate of the car in front of me at highway following distances during day. This is simply a limitation of the sensor and processing power of the chipset.

3. Forget about being able to capture plates at night unless there is no speed delta between you and the other car. It is shutter speed that determines motion blur (not frame rate), and there is no chance that 1/30 or 1/60 is going to pick up plates on a car that is blowing pass you at 20MPH over your speed. It's always a good idea to have the mic on and call out the license plate if you are able to.

 

[andrew124C41]

1. Blackvue cameras are not known for their image quality. People who buy them because they like the design, features and brand value. There are multiple threads on this forum complaining about image quality of their flagship DR900S.

2. 1080p simply doesn't have the resolution to capture plates from more than ~25 feet away with a wide angle lens. I have a Viofo A129 Pro now, and even it has problems reading the plate of the car in front of me at highway following distances during day. This is simply a limitation of the sensor and processing power of the chipset.

3. Forget about being able to capture plates at night unless there is no speed delta between you and the other car. It is shutter speed that determines motion blur (not frame rate), and there is no chance that 1/30 or 1/60 is going to pick up plates on a car that is blowing pass you at 20MPH over your speed. It's always a good idea to have the mic on and call out the license plate if you are able to.

Unfortunately, you are absolutely correct. I did not have time to research this fully. I had just purchased a brand new Outback, fully loaded for my wife. We live in the DC area and drivers are very rude and selfish. Less than a month after buying her car someone backed into it on our parking lot. So, she really wanted a dash cam. I paid something like 300 for the Blackvue thinking that surely that should be enough for a good dashcam.

That it could not read a plate at night from the car right behind us really pissed me off.

Blackvue got all these great ratings but frankly I think they are over priced and falsely advertised.

I just purchased a new Lexus for myself and wanted a dashcam. I probably will not buy one at this time...not until they can read a plate. I will just depend upon our insurance.

A camera needs a sensor capable of low light capture as well as a frame rate faster than 60, perhaps 120 might do it...I don't know for certain.

I do suspect my wife's iPhone could work but I have not as yet tested it.

Bottom line, skip the dashcam and get good insurance unless you are an Uber driver..then use one for the interior to protect yourself from radical feminist predetors.

 

[kamkar]

Many people still have too high expectations of what a dashcam can do, both in regard to small details like a plate, or in general in relation to the footage.

Plate capture are only icing on the cake, you should not expect to get those all the time, you only stand a decent chance if it is daytime and a pretty clear and sunny day.
Actually light levels dont have to drop muuch before dashcams start to struggle, and a thing like plate capture at speed become impossible, just driving into shade on a sunny day ( trees lining road on either side ) and you are in problems in regard to small details getting captured.
If you expect to capture plates at driving speeds at night you probably have to wait 10 years before getting a camera, probably even longer.
Do remember plates are only needed if you experience a hit and run, otherwise i think it will be pretty easy by your camera footage or the scene of the accident to determine what happened.
And the great picture, even pretty poor dashcams will be able to deliver ( your positioning in relation to lane markings or side of the road / color of intersections / speed even without GPS )
All these simple and larger things a dashcam will do for you will be able to prove the most outrageous claims against you.
As long as you can prove what you did / did not do, you should be safe from most claims against you, remember you have some evidence, the other part just have words.

Dashcams are not cinematic devises, it is a camera alright but if you want to record scenic drives ASO you are probably better off with another kind of camera, and one with a more narrow field of view.

 

[comadose]

Forget about being able to capture plates at night unless there is no speed delta between you and the other car. It is shutter speed that determines motion blur (not frame rate), and there is no chance that 1/30 or 1/60 is going to pick up plates on a car that is blowing pass you at 20MPH over your speed.

Yet it works fine with cars going past a speed camera at 100+ MPH delta.

Remember that licence plates are reflective and lit by the car as well as your own lights. The issue is usually that the camera sets the exposure so that most of the scene is visible and the licence plate gets blown out as pure white and blurred by low shutter speed. If they used a sensor with bigger pixels (more light per pixel) and better HDR they could capture it.

 

[relive]

Yet it works fine with cars going past a speed camera at 100+ MPH delta.

Remember that licence plates are reflective and lit by the car as well as your own lights. The issue is usually that the camera sets the exposure so that most of the scene is visible and the licence plate gets blown out as pure white and blurred by low shutter speed. If they used a sensor with bigger pixels (more light per pixel) and better HDR they could capture it.

Flash photography, plus a DSLR inside. This is not remotely the same thing as a dashcam. Flash durations can be as short as 1/10000s for the power output that a speed camera requires at night. You can read a plate off the Thrust SSC with that flash duration. Unless you plan on mounting a DSLR and a strobe to your windshield, please stop comparing that to a dashcam.

 

[comadose]

Flash photography, plus a DSLR inside. This is not remotely the same thing as a dashcam. Flash durations can be as short as 1/10000s for the power output that a speed camera requires at night. You can read a plate off the Thrust SSC with that flash duration. Unless you plan on mounting a DSLR and a strobe to your windshield, please stop comparing that to a dashcam.

The newer ones don't flash. It distracts drivers and has caused accidents. It's not needed anyway, just browse some photos of cars at Le Mans at night.

Yes, it's a DSLR. That's the point - if you made a dashcam with a bigger sensor it could do it. But they keep using phone camera sensors instead.

 

[Nigel]

You can read a plate off the Thrust SSC with that flash duration.

Might even catch the Bloodhound LSR at the moment, still a long way off top off their 1000MPH target, and I noticed they are having a few issues with their rear dashcam - those magenta flashes aren't right, reminds me of the early mini 0806!

 

[relive]

Yes, it's a DSLR. That's the point - if you made a dashcam with a bigger sensor it could do it. But they keep using phone camera sensors instead

No, it seems like you missed the point entirely. Still photography cannot be compared to video. I am a commercial photographer, and when I work alongside film crews, I can capture way more detail and dynamic range in stills with my "measly" $3k cameras than they can with their $40k 8K RED cameras. The reason is exceedingly obvious: still cameras are only required to shoot, on average, one frame every few seconds, while video cameras need to shoot continuously at at least 24fps. Video cameras need to use highly lossy codecs and chroma sub-sampling to deal with the massive amount of data. While the 58Mb/s bitrate of the A129 Pro front camera is considered high by dashcam standards, uncompressed 4K30P video is 5,304Mb/s, so 58Mb/s represents a nearly 100:1 compression ratio.

You seem to also not understand the physics of lenses and the relationship between sensor size and depth of field. Yes, it's possible to achieve around a 1/500s shutter speed in a poorly lit scene at night with a full-frame DSLR sensor. However, the Le Mans pictures that you refer to were taken with a 400mm f/2.8 (or similar lens) at f/2.8. Take a look at the background and tell me how much detail you can see other than the single car that is in focus. When you increase your sensor size, you need to stop down your aperture to maintain the same depth of field, which then prevents you from achieving the required shutter speed to freeze action. This is like adding more fuel to a rocket to increase your delta-V, but the weight from the extra fuel reduces your delta-V, causing you to have to add more fuel, etc, etc.

Might even catch the Bloodhound LSR at the moment, still a long way off top off their 1000MPH target, and I noticed they are having a few issues with their rear dashcam - those magenta flashes aren't right, reminds me of the early mini 0806!

The g-forces must be insane. I wonder much drag there must be at 1000MPH, even with the tiny Cd of the vehicle.

 

[Nigel]

The g-forces must be insane. I wonder much drag there must be at 1000MPH, even with the tiny Cd of the vehicle.

20 tonnes of drag at 1000 MPH (1610 Km/h), 135,000 bhp.
"At 1,000mph, the wheels experience 50,000 times the force of gravity trying to tear the wheel rim apart, "

Not just aerodynamic drag, it runs several mm below the surface so there is quite significant ground drag at 1000MPH!

I was surprised how much steering effort it takes, today's test run:

 

[comadose]

You seem to have missed my earlier comment that I'd prefer a lower frame rate, i.e. more like a series of stills.

Something like a micro 4/3 sensor and optics would be good.

 

[kamkar]

I have said i would even take B&W footage if that ensure i will get more detail at night, and i have been playing with the idea of a dashcam aided by IR light on the front of my car.
but just adding a lot of light may not ensure plate capture, for the reason mentioned elsewhere already.
BUT ! i pretty much never think of plate capture, it only interest me when i do testing on new dashcams, and then its really just a way to conclude the camera are not totally garbage.

 

[comadose]

I have said i would even take B&W footage if that ensure i will get more detail at night, and i have been playing with the idea of a dashcam aided by IR light on the front of my car.
but just adding a lot of light may not ensure plate capture, for the reason mentioned elsewhere already.
BUT ! i pretty much never think of plate capture, it only interest me when i do testing on new dashcams, and then its really just a way to conclude the camera are not totally garbage.

I don't think IR will help much because the plate is likely to be uniform temperature.

The problem is that they use small camera phone sensors that don't get a lot of light, and then process the image to give an "HDR" effect where the main scene is visible but the number plate is just a blurred, blown out blob. I think a high end small form factor sensor tuned just for licence plate capture might work, but I don't have a dashcam with enough control over the image to try it. Certainly if you get an older/cheaper phone and use a camera app with manual exposure settings you can get an image that shows the plate with a bit of post-processing to enhance the contrast.

 

[kamkar]

Plate cameras ( CCTV ) do a fine job with a IR flash, but they are probably also tuned a lot different from a regular CCTV camera or a dash camera.
The Danish police cars with ANPR can so plate captures at speed, so that's 2X the challenge compared to a stationary ANPR system like a fixed speed trap.
But when you look at the footage of such systems it is also clear to see, that the plate are the main thing, and not so much the car or the person inside it.

Both day and night it is evident the Danish mobile speed traps use a high shutter speed, by the dark pictures even in daylight, and so the use for a good flash at night.

I just wanted to try IR aided cameras to capture more of the surroundings, in regard to plate capture it will probably be worse due to the reflective plates and me using a camera to see more in general and not plates.

I think you confuse IR with thermal cameras "likely to be uniform temperature"
Though heat are also light, then you have to go even lower than the normal IR 850 - 950 nano meter wavelengths.

But in general yes bigger sensors are better assuming they are bigger cuz the pixels on it are bigger, +100 megapixel phone cameras are just around the corner. but in a non binned format that got to be garbage for low light photography of things with some degree of movement.
But bin the pixels to 9 in 1 and you will probably have fine 10 mpix pictures.

But while thermal cameras have become a lot more affordable, the cheaper models are still a very low resolution.

There is a clear difference using my phone camera as its native 48 mpix VS using the same sensor and settings but with a 4 in 1 binning to get a 12 mpix picture, a 4 X larger pixel do wonders.

 

[comadose]

IR flash won't blind the driver at least. Seems like a decent solution for fixed cameras but obviously not for dashcams.

 

[SawMaster]

It's good to remember that dashcams are essentially low-end consumer grade cams trying to be good at doing many different things at once. To get best possible plate capture will necessarily result in otherwise lesser videos in some way, and that can be important too. I think the best approach is to have multiple cams so that one can be 'tuned' for plate capture while the other one(s) handle the rest.

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. 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.

Phil