Nighttime Front vs. Rear Plates- Should I expect better?

[Swampledge]

After several incidents in our neighborhood, I decided it’d be better to capture plates of all vehicles passing, rather than those just entering my driveways. The below pic is essentially where I’m at, although the front plates currently look a bit cleaner after I reduced the sharpness. Camera is a IPC-HFW5241E-Z12E at max zoom. Shutter is at 1/2000 since vehicles are traveling at about 45 mph. This is a rural area, so there is no external lighting. The challenge I’m having is the approaching front plates show up much brighter from the IR of my camera than the illuminated rear plates of the cars traveling in the opposite direction. This photo shows a comparison of both as two cars passed within the same frame.

Just looking for the opinion of you more experienced guys with this if I should expect to be able to do better. FWIW, I’ll soon be adding a second camera looking in the opposite direction to get two shots of every vehicle, since so many lack front plates, despite being required to by law.

 

[wittaj]

There are so many factors that come into play. Angle of camera to streetm height of plate in relation to camera, etc. all come into play with how the light is reflected off a plate and back to the camera. Location of plate in the field of view can change it greatly as well.

Many of us drop sharpness down below 10 - have you tried that?

I have been doing this long enough to know that it will all average out over time. I will get a car go by and the front plate is way better than the rear plate and vice versa.

Adding the 2nd camera will help as well.

If you try to chase perfection and get 100% of the plates, you will make it worse LOL. There will always be that problematic plate.

 

[Swampledge]

Thanks, wittaj. Dropping the sharpness that low will be my next tweek, along with inching the contrast up some more. My sharpness is currently around 30, and contrast high 70’s. I’ve actually become disciplined enough that I’m keeping a log of my changes along with photos like I posted so I can compare and assess my progress. At some point I’ll decide I have done all that I can- I’m just doing an early assessment of how close I am to that point.

I’m very grateful to you and the others here, who have posted so much insight that got me to where I am with this so quickly.

 

[bigredfish]

Whats the distance to the target?

I'm guessing the far plate is simply not getting enough IR. Possibly simply due to angle and distance.

I capture at 90-100ft with these settings though there is some ambient light from street lights

 

[trapper]

Every time I see a post like this it causes my to question my current LPR settings thinking I can get better. like mentioned above I end up disappointed and back to my original readable pics. Good is good,

 

[Swampledge]

Thanks, bigredfish. I see that you’re able to use 1/1000 shutter speed. I‘m planning to try that to see if I can get away with it, but it seems like maybe your location has lower vehicle speeds. We’re on a straight stretch of state highway with good pavement, so….

I think my distance is closer to 150 feet. The camera is camouflaged in landscape shrubs, and I was looking to get as little angle as possible. When I ran the cable, I also ran a low voltage cable to be able to power a separate IR illuminator. Not so much to provide brighter IR, but so as to not attract spiders and bugs toward the camera itself.

As I mentioned in my response to wittaj, I’m very familiar with the adage, “If it ain’t broke, fix it ‘til it is.” That’s something I know how to do. I’ve got a few changes yet to make, and then repeat with the 2nd camera.

 

[Mike A.]

Yeah, you can play with this forever. You get to a point where you make one aspect slightly better and you make another not quite as good when things change in some way. Need to find a happy medium somewhere and call it good enough and just step away.

 

[Parley]

I have 4 LPR cameras. The shutter speeds are 1/500, 1/1000, 1/2000 and 1/4000. I would never have dreamed it would turn out that way. The cameras are all Dahua 5442E-Z4E models.

 

[bigredfish]

Yeah agreed.
I have 6 set at anywhere from 250-4000 and @Swampledge youre correct, I see speeds of 25-35 at that location so 1/2000 or more would be better for you in theory. I run a fixed shutter

Try playing with the Iris control to see if that helps with the “dot” effect…that close plate looks good otherwise…

 

[tangent]

You could try experimenting with a separate IR illuminator.

 

[Parley]

You could try experimenting with a separate IR illuminator.

A separate IR illuminator helped in my situation. In fact, I have purchased a 2nd one for another location.

 

[jeremycolvin]

@Swampledge I am using the newer IPC-HFW5241E-Z12E-S2 at about 200 feet or so, just the cams, no extra ir illuminator and here's what I'm getting in what sounds like a similar angle to you, though mine may be a little more straight on as mine are literally on my mailbox... but I would say "yes" I think you can get better (assuming that not much has changed from the IPC-HFW5241E-Z12E to the IPC-HFW5241E-Z12E-S2. - Traffic through here goes about 25-35ish mph

 

[Swampledge]

Thanks, Jeremy. I just finished my tweaks for this evening, and think I’ve made significant progress. Nothing worth posting yet, ‘cuz traffic has gotten pretty light, and so I don’t have a good sample size. But I found I could get away with slowing my shutter to 1/1000 and get a brighter image on the rear plate. I also dropped my sharpness a good bit, which cleaned up the image, and reduced my Iris too. (No, I didn’t shotgun them. I made changes one at a time, collected images, reviewed them, then made additional changes.) Now I’m just waiting to accumulate images overnight for review in the morning.

 

[Swampledge]

After reviewing a bunch of images from overnight, I'm on the right track. Still got a ways to go to be happy. It was wet last night, so the rear plate images, which are the ones I'm most concerned about improving, were sometimes I think obscured by spray from the vehicle's rear tires. And, as many of you know, there can be huge variability from easily read to unreadable from vehicle-to-vehicle with the same camera settings. The good news, though, is that almost 100% of the front plates are readable. The main thing, though, is I have a plan for the parameters I plan to mess with tonight. Then there'll likely be a new plan for the next night. I realize I haven't said it, but daytime captures are a cakewalk; no challenges for my configuration there.

Thanks again to everyone who's offered suggestions.

 

[jeremycolvin]

After reviewing a bunch of images from overnight, I'm on the right track. Still got a ways to go to be happy. It was wet last night, so the rear plate images, which are the ones I'm most concerned about improving, were sometimes I think obscured by spray from the vehicle's rear tires. And, as many of you know, there can be huge variability from easily read to unreadable from vehicle-to-vehicle with the same camera settings. The good news, though, is that almost 100% of the front plates are readable. The main thing, though, is I have a plan for the parameters I plan to mess with tonight. Then there'll likely be a new plan for the next night. I realize I haven't said it, but daytime captures are a cakewalk; no challenges for my configuration there.

Thanks again to everyone who's offered suggestions.

I'm sure that you've already done this... but for me the MOST HELPFUL thing for tweaking my cam settings in was to have my wife go in her jeep to my main focal point and just sit there in the jeep so I could tone it in... if someone came and she needed to move then she moved and would go back to that point for me to continue tweaking... saved a ton of time.

 

[Swampledge]

I'm sure that you've already done this... but for me the MOST HELPFUL thing for tweaking my cam settings in was to have my wife go in her jeep to my main focal point and just sit there in the jeep so I could tone it in... if someone came and she needed to move then she moved and would go back to that point for me to continue tweaking... saved a ton of time.

Thanks, but that’s not happening. Straight road, state highway, smooth pavement, so high speeds. I sit on our front porch and am disgusted by the number of drivers that are looking at a phone instead of the road as they go by. Includes tractor-trailers! Too risky for people and our vehicles. I can stick a license plate on a stick out there, but that doesn’t simulate the variety of rear plate lighting situations. As I noted, I’ve got some clean rear plate captures and complete misses with the same camera settings. I just need to develop a feel for what percentage is reasonably attainable.

Keep in mind, I’ve only had a use for this information about twice per year over the last few years, and only once was it for me. The rest of the time it’s been to help neighbors.

 

[The Automation Guy]

This is to no one in particular, but just a reminder to all....

Keep in mind that your settings are going to be unique for your situation. How/where you cameras are mounted can/will play a role in what settings work best. For example, if your camera is set up more parallel to the street (ie you are shooting almost straight down the street), your shutter speed can be lower than a camera mounted farther back from the street that is mounted more perpendicular to the street. This is because an object moving directly at the camera "moves" less than an object moving across the field of view. So while someone might need a 1/4000th shutter speed because their plates are moving across the field of view, someone else might easily get away with 1/1000th because they are positioned near the road and the field of view looks down the road. Other variables like traffic speed, ambient light levels, etc play a huge role too.

So while looking at settings of other users can be EXTREMELY helpful (and I appreciate everyone that takes the time to post their settings), don't be afraid to experiment as see what works best for you!

 

[jeremycolvin]

This is to no one in particular, but just a reminder to all....

Keep in mind that your settings are going to be unique for your situation. How/where you cameras are mounted can/will play a role in what settings work best. For example, if your camera is set up more parallel to the street (ie you are shooting almost straight down the street), your shutter speed can be lower than a camera mounted farther back from the street that is mounted more perpendicular to the street. This is because an object moving directly at the camera "moves" less than an object moving across the field of view. So while someone might need a 1/4000th shutter speed because their plates are moving across the field of view, someone else might easily get away with 1/1000th because they are positioned near the road and the field of view looks down the road. Other variables like traffic speed, ambient light levels, etc play a huge role too.

So while looking at settings of other users can be EXTREMELY helpful (and I appreciate everyone that takes the time to post their settings), don't be afraid to experiment as see what works best for you!

Absolutely That's the only reason that I posted my settings, along with setup (cam to road) diagram... it sounded to me like we had similar situation/set up - other than it sounds like is traffic moves a bit faster. I will add 1 general note too... I thought that having a nearly straight on, would make it a no brainer / best set up... and it DOES have it's advantages (longer time/opportunity to capture plates, but ALSO has it's disadvantages.. namely target zone. Setting up the trigger zones for this set up has been nothing short of a pain in the butt. Personally I would not recommend the set up I'm using to anyone. (Mailbox height, cams on mailbox, shooting down a straight road). With trees, etc in the background it's next to impossible not to have a million false triggers, because you have to have cams pointed high enough to get plates, but in doing that you also have all the trees, etc in the background... plus, no matter how you set your trigger zones up, you start triggering WAY TOO SOON.... so to combat that you have to lower trigger times to make it keep triggering every second until it finds a plate in the view... I'm thinking about installing a flag pole or something next to the mailbox so I could get the cams up about 8-10 feet, so I could angle them down instead of straight, etc... point here being... what I've learned in doing this setup is that there are 3 main and VERY important things to pre-plan for...

1) Good Cams that will work well in darkness of night with good to great zoom capabilities. (I used IPC-HFW5241E-Z12E-S2 - which have been great)

2) Location, Location, Location... as stated above, try to plan for premium placement, think about the background, the target zone, the time in shot, speed of traffic, etc... I can't say this enough.. pay attention to your background (what's behind the area that you anticipate capturing the cars in)

3) Patience... it can be a very frustrating journey to get dialed in correctly... go into it knowing that it won't be "easy" - but it's worth it in the end.

 

[Swampledge]

You've got great captures, Jeremy. I'm learning a lot, one of which is the need for patience or tenacity. I'm not looking to use AI to capture plate numbers; I just want to be able to pull one one up if needed. I'm not even sure how much I'd trust AI to OCR the plate numbers. (As a side note, two nights ago CPAI told me there was a person on our back deck. Upon inspection, it was a possum. At least it didn't identify me as one) Anyway, because reading plate nos. in the day has been so easy, I haven't bothered looking at too many. But I spent some time doing that this morning, and found that I occasionally have to look at several frames to differentiate between B's and 8's, and sometimes between D's and 0's, etc. I did find it beneficial to run the LP cam at 20 fps vs. my usual 15 to get an extra frame or more to compare. Looking at plates in the daytime also helped me understand why some plates are much harder toread than others. there's a lot of bent plates going by.

You also mentioned that having a more straight on view gave you challenges with triggers. I really haven't had any difficulty with triggers; I'm using no zones, simple algorithm, 0 sec duration, and about 2 seconds pre-trigger. My FOV is so tight, passing cars pretty much fill it, and the background really has nothing to cause triggers when there's no traffic-especially at night.

I did find when I inadvertently changed my camera angle to a sharper one that it seemed to even up the far lane rear plate brightness at night to be a bit closer to the front plates, but I think that might be more related to the plate being closer to the camera. I've since readjusted the camera, so I'll see what happen tonight.

 

[jeremycolvin]

You've got great captures, Jeremy. I'm learning a lot, one of which is the need for patience or tenacity. I'm not looking to use AI to capture plate numbers; I just want to be able to pull one one up if needed. I'm not even sure how much I'd trust AI to OCR the plate numbers. (As a side note, two nights ago CPAI told me there was a person on our back deck. Upon inspection, it was a possum. At least it didn't identify me as one) Anyway, because reading plate nos. in the day has been so easy, I haven't bothered looking at too many. But I spent some time doing that this morning, and found that I occasionally have to look at several frames to differentiate between B's and 8's, and sometimes between D's and 0's, etc. I did find it beneficial to run the LP cam at 20 fps vs. my usual 15 to get an extra frame or more to compare. Looking at plates in the daytime also helped me understand why some plates are much harder toread than others. there's a lot of bent plates going by.

You also mentioned that having a more straight on view gave you challenges with triggers. I really haven't had any difficulty with triggers; I'm using no zones, simple algorithm, 0 sec duration, and about 2 seconds pre-trigger. My FOV is so tight, passing cars pretty much fill it, and the background really has nothing to cause triggers when there's no traffic-especially at night.

I did find when I inadvertently changed my camera angle to a sharper one that it seemed to even up the far lane rear plate brightness at night to be a bit closer to the front plates, but I think that might be more related to the plate being closer to the camera. I've since readjusted the camera, so I'll see what happen tonight.

Nice! Sounds like you are definitely on the right track. And I completely agree with your assessment to "not trust OCR" - It's definitely NOT a perfect science. What I do like about using LPR is that it "alerts" on the pates and even though the OCR may read it wrong, it makes it MUCH faster to investigate the video clips to find a good zoomed in image of the plate by CTRL+Double clicking on the alert which brings you right to the frame that it thinks it sees a plate on. NOTHING beats the human eye. For me, I record on those 2 cams 24 hours a day and use the LPR to alert and set queues on the vid clips that it finds the plates on to look at the actual plate with my own human eyes