Subscriber Discussion

Vehicle Break In, Unable To Identify Plates

UI
Undisclosed Integrator #1
Jul 30, 2017

We had a vehicle break in at a customer's site. The camera was able to get decent footage of the actual break in but unfortunately not even close to identifying the plates. We want to resolve this for future incidents. 

Just some specs on the camera:

It's a 2.8mm lens with a resolution of 1920x1080 -- It's HikVision if that matters. 

I'm thinking of adding some external IR illuminators, I am not sure what else needs to be put in place to identify the plates. Do we need to put IR / replace the camera to a higher MP camera? 

What do you guys think?

(1)
UI
Undisclosed Integrator #2
Jul 30, 2017

The 2.8 lens is probably too wide to capture plates. Without seeing the site my recommendation would be to add another camera with a narrow field of view on the entry/exit point of the property. 

(3)
JH
John Honovich
Jul 30, 2017
IPVM

What's the model number? This will let us verify the exact AoV.

How far was the car from the camera? This will let us verify the exact pixel density delivered.

That noted, I agree with #2. The FoV is probably too wide to capture plates. With a 2.8mm lens, the HAoV will likely be ~100°. Here is an example of what that looks like with a car 50' away:

Even at 25' away it is going to be challenging in best case conditions:

You can try out distances / resolution / AoV with this sample calculation.

Given your questions about IR, I assume this happened at night. More IR might help but we first need to verify pixel density, e.g., if the car is 25' away, this means around 30-35ppf which will be vey hard to get license plate details at night even with 'perfect' IR.

(4)
JH
Jay Hobdy
Jul 30, 2017
IPVMU Certified

 

The top is the customers camera, the bottom is a Speco O2LPR67 license plate camera. Same location, same scene, cameras are mounted on the same pole.

 

I really do not see the FOV being the issue, I think it is going to be more of a lighting/reflection/exposure issue.

 As you can see from the video, as the cars drive out, headlights shine on the tags at certain times, and here in GA tags are reflective. This creates havoc on our regular cameras.

We have set up quite a few license plate cameras. Some as LPR coupled with Milestone, and some as just LPC, being able to read the tag during playback. We always go for a chokepoint, like an entrance, then have the camera focused on that area.

 

We have a general starting point for our settings such as IR and exposure, but have to tweak certain sites due to various environmental issues.

 

The Speco is around $500 and we are now using a Uniview that is around $400 dealer cost. We used Uniview but due to manual focus and limited focal range (22MM) we tried the Specos. The Specos have had an issue crashing IE, and have these teeth on their adjustments which makes detailed adjustments impossible. The new Univiews have motorized lens up to 47 MM so we went back to them.

 

 

(1)
(1)
JH
John Honovich
Jul 30, 2017
IPVM

The top is the customers camera, the bottom is a Speco O2LPR67 license plate camera.

I really do not see the FOV being the issue

But in the two videos you showed, the LPC camera has a much narrower FoV, certainly that helps, no? 

Images comparison:

Also, in your example, the LPC camera overall light levels were very low, which would likely cause images trying to see other details like people.

(1)
JH
Jay Hobdy
Jul 30, 2017
IPVMU Certified

The top camera is a Ubiquiti camera, which the customer installed and still specifies... I can tell you at 25+ properties, those Ubiquiti camera can not capture license plates from 10' at night. All you get is a white blur. I have seen the same thing with multiple Dahua cameras.

When we install the LPC/LPR cameras, we always have a second camera for overview. We are VERY upfront in the sales process that our tag cameras are there for one purpose, to capture tags. People, overview etc needs to happen from a 2nd camera.

 

IME it is not the resolution, it is the brake lights, license plate lights, IR bouncing off the tag that creates the problem.

 

This camera was installed by another company, and I know it is a Dahua 3MP turret, not sure which model. As you can see FOV has no real relevance here as you can not come close to reading the tag.

(1)
(1)
JH
John Honovich
Jul 30, 2017
IPVM

IME it is not the resolution, it is the brake lights, license plate lights, IR bouncing off the tag that creates the problem.

Jay, I agree with you that brake lights, license plate lights and IR bouncing off the tag creates problems. However, in the first video above, the resolution is a contributing factor as well. That's a very wide FoV and even in perfect day time conditions reading the license plate is going to be hard for a 1080p or even 3MP camera.

When we install the LPC/LPR cameras, we always have a second camera for overview.

And I think that is a good design.

However, you can get license plates with regular IR cameras at night, not always but in many conditions, by making the shutter much faster, e.g., Low Cost License Plate Capture ShootoutLicense Plate Capture Shootout. Note: the shutter change is needed for moving vehicles, if all one wants is parked vehicles, which OP may be implying here, fast shutter is not needed.

(1)
JH
Jay Hobdy
Jul 30, 2017
IPVMU Certified

And this is where we probably need to take the camera class. I really wanted to get on this session but just could not commit the time.

 

We are using shutter speed to darken the image. If we see a tag at night and the tag is still too bright, we will choose a faster shutter speed, or a faster range.

 

Sometimes we adjust the IR as well.

I have heard some have had good success with regular cameras. The odd thing is when I asked my Dahua OEM, and Dahua direct, about a LPR/LPC camera, nobody suggested adjust shutter. They just said they had no solution....

(1)
(1)
(1)
JH
John Honovich
Jul 30, 2017
IPVM

The odd thing is when I asked my Dahua OEM, and Dahua direct, about a LPR/LPC camera, nobody suggested adjust shutter. They just said they had no solution....

Well, that does not say much for them.

Btw, a fast shutter speed is typically pre-configured / forced in LPC and LPR cameras. As an example, see our Geovision LPC test report, where we note:

Due to its fast exposure (1/500s slowest) required for reliable plate capture, the GV-Hybrid 10R is not suitable for general surveillance purposes due to much darker images than typical.

(1)
(1)
UI
Undisclosed Integrator #1
Jul 30, 2017

Jay,

Do you use the Speco/Univiews for both capture and through the Milestone LPR? Or do you have other cameras that you use for ALPR.

How are the specos and univiews with cars at a faster speed? 

I think adding a secondary camera for LPR at entrances is a necessity at this point. We need general overviews + LPC. 

(1)
JH
Jay Hobdy
Jul 30, 2017
IPVMU Certified

Yes we have used the Speco and Uniview with Milestone as ALPR, putting tags into a database. The "after" video above is connected to Milestone license plate program, and after a learning curve, the cameras seem to be working great.

 

Otherwise we use them with Dahua NVR's and they work great there as regular cameras. No database but if something happens and they want to be able to read a license plate, they can.

 

Key things to remember if using Milestone:

Set cameras to MJPEG

4-5 FPS

Learn how to take snapshots and not use the liveview for tag samples

Set the cameras up with as parallel shot as possible

 

As far as speed, we use these at gates, so speed has never been an issue. But the exposure on these is set extremely fast, so there should be no blur. The bigger challenge may be the 4-5 FPS and the field of view. It is possible the car may not even get recorded unless you are able to keep it view for enough time.

 

 

(1)
(1)
UI
Undisclosed Integrator #3
Jul 31, 2017

If you don't mind me asking, who is your Uniview supplier? Just been wanting to check it out.

JH
Jay Hobdy
Jul 31, 2017
IPVMU Certified

Safe Cam out of Orlando FL. 407-259-2683.

 

There are are only 3 US distributors that I know of that stock the IPC262ER9-X10DU. The other is Gen IV. They are higher priced but may offer better support.

 

Make sure you get the set up instructions which shows you which default settings to change. Or I can email it. 

UI
Undisclosed Integrator #1
Jul 30, 2017

Jay. This has been extremely helpful. 

Besides exposure and using MJPEG (For milestone) are there any other settings that need to be changed out of the box to capture good plates? 

Did the Speco need external IR illuminators, or that is all with the built in illuminators?

Can the Speco capture 2 lanes of cars? I know with some LPR cameras since the FOV is so narrow it can capture about 6' wide

 

(1)
JH
Jay Hobdy
Jul 30, 2017
IPVMU Certified

In the two videos, the Speco replaced a standalone IR.

 

We have not used any external IR at all on any LPR/LPC cameras.

 

The Speco rep said our FOV was too wide on that video. Not sure how he made that determination, as Milestone doesn't care as long as everything meets their requirements with character height, angles etc. Milestone can read multiple lanes, I believe they need a certain pixel per foot though.

 

The challenge we have is in apartment complexes, drivers tend to drive in the middle of the road unless there is oncoming traffic, so we have to cover a wider area.

 

I would suggest setting one of these up at your shop, somewhere close, or have remote access to the camera like VPN or remote desktop.

 

I have spent a great deal of hours configuring these 10 Milestone sites. I am spending my Sunday logging into 10 sites, checking settings, making sure Milestone is reading tags, adjusting exposures etc.

 

A good part was we were very green to it, and we should have all these sites tweaked and finished after a site visit to reaim a couple cameras.

(1)
Avatar
Amer Alsamarraie
Jul 30, 2017

To identify the number plate, you need the following :

1- The camera should be at least 1.2 up to 3MP.

2- If the light condition is poor, you have to use IR source or white LED (if you want to identify the plate color). However, if the car is moving, you should have pulsed light source.

3- You have to choose a right LPR/ANPR engine, which should have been already trained for the country's number plate where the engine is installed.

4- The installation is very important aspect. The distance between the loop and the camera should be well calculated and well synchronized between the shutter time and the light source.

5- The characters (numeric or alphabetic) on the captured image have to be at least 20-23 pixel high.

Amer Alsamarraie

(1)
JG
Jeff Gack
Jul 31, 2017
IPVMU Certified

I have had the best luck getting plates by:

1. Using 2 cameras, 1 for an overview with wide FOV and setup for low light. The other with a narrow FOV mounted at a good angle to the plate.

2. Camera aimed at a choke point, usually the exit where they stop to pull out or where they're going the slowest.

3. Camera with IR illumination and a varifocal lens to get the narrow AOV.

4. Cameras shutter speed set fast for night usually around 1/200th and then adjust the gain to properly expose the plate. (in a dark scene, this will usually give a black picture until the car drives by) The customer will think this black camera at night is a waste until you explain the importance of getting that plate.

5. Focus camera at night to get good focus.

(1)
(1)
Avatar
Jon Dillabaugh
Jul 31, 2017
Pro Focus LLC

I didn't read all of the replies, so if this is redundant info, please forgive me.

You have multiple issues here.

1) FOV needs to be as narrow as possible. This will ensure that you get as many PPF as possible. Cover the lanes only. Try to pick the narrowest area (pinch point). Varifocal cameras are usually the best choice here as they offer more flexibility.

2) Angle of incident needs to be as straight as possible. Make sure that vehicles aren't turning in the area you choose to cover. The less angle, the better.

3) Lighting conditions need to be ideal and your camera must be adjusted to overcome headlights and taillights. Most decent cameras can be adjusted to match the scene. You will need to greatly change the minimum exposure rate. Each camera and scenario will be unique. No two sites will be identical. This will take trial and error to get perfect.

4) Vehicle speed will also be a possible issue. If you have fast moving cars, you will likely need higher frame rates to get enough frames of the object in scene. The more frames you have to choose from, the more likely you are to get one that is clear enough.

(1)
New discussion

Ask questions and get answers to your physical security questions from IPVM team members and fellow subscribers.

Newest discussions