Blue Iris and CodeProject.AI ALPR

[OfficerD00fy]

with only 12 + post-trigger images I seem to be missing vehicles, especially if they are driving fast. I've increased the +post trigger images to 40 with 50ms.
Or should I increase the analyze to 100ms?
Also added+ pre-trigger images to 10.

 

[Tygunn]

with only 12 + post-trigger images I seem to be missing vehicles, especially if they are driving fast. I've increased the +post trigger images to 40 with 50ms.
Or should I increase the analyze to 100ms?
Also added+ pre-trigger images to 10.

How long is your system taking to process the 50 images you're sending to CPAI?

As Actran pointed out to me, if the AI is taking too long to process, you may be losing frames in the processing pipeline. The BI manual actually covers this; I bumped up my pre-trigger record time and found things were more reliable.


If you step through individual frames of a capture for a car moving by quickly you'll be able to get an idea of how far a car can move in X amount of time. 40 post trigger images 50msec apart is 2 seconds, so 20fps that you are processing. I think it would be reasonable to increase the spacing to 100ms and lowering to 20 images.

Make sure to save the AI analysis and load them up in the AI tab of BI status to see how the analysis is going.

 

[OfficerD00fy]

How long is your system taking to process the 50 images you're sending to CPAI?

As Actran pointed out to me, if the AI is taking too long to process, you may be losing frames in the processing pipeline. The BI manual actually covers this; I bumped up my pre-trigger record time and found things were more reliable.


If you step through individual frames of a capture for a car moving by quickly you'll be able to get an idea of how far a car can move in X amount of time. 40 post trigger images 50msec apart is 2 seconds, so 20fps that you are processing. I think it would be reasonable to increase the spacing to 100ms and lowering to 20 images.

Make sure to save the AI analysis and load them up in the AI tab of BI status to see how the analysis is going.

This is my AI detection rate. Trying 10 pre record triggers and 20 post trigger records at 50ms. Found out I was BL set the camera to 60 FPS when camera was running 50 FPS. Will see if it fixes it?
A car passing takes about 1 second...
It says 7seconds to process.

 

[OfficerD00fy]

Seems to be working better however I noticed one vehicle slip pass when the AI was doing a detection. Can it only run one at a time? Another car came through just as it was finishing the AI detection and the AI missed it. It seems if something else triggers the camera trigger then it will miss the vehicle...

I think this has to do with the break time.
It was
end trigger unless re-triggered within 5.0 sec
maximum trigger/alert duration 20 sec.

I've modified it to
end trigger unless re-triggered within 1.0 sec
maximum trigger/alert duration 2 sec.

Would that fix it?

 

[JNDATHP]

Any suggestions to improve OCR accuracy?

Using a HFW5241E-Z12E at maximum zoom. Height is approximately 11’ (used an existing hole from previous owner) and distance is approximately 140’.

The plate reads 15B KGZ but OCR reads 1583 KGZ.

BI 5.8.4.5 and CPAI 2.6.2.0 and LPR module 3.1.0

 

[Vettester]

Any suggestions to improve OCR accuracy?

To improve accuracy you'll need to decrease the camera angle.

 

[MarioBoo]

Yeah that’s a crazy angle to work with.

 

[JNDATHP]

So more distance?

 

[MikeLud1]

Any suggestions to improve OCR accuracy?

I am working on adding the option to save OCR training images then users can send me the training images so I can retrain the OCR model to improve accuracy. Most likely I will release a new version in a month or two.

 

[wittaj]

So more distance?

It's a tradeoff then, especially at night.

I have seen that it isn't super accurate. It will get more accurate as time goes on, but even with a good straight on shot mine will miss as well.

 

[Vettester]

So more distance?

If possible move the camera closer to the street. Mine are about 15' from the street so I get good shots in both directions.

 

[OfficerD00fy]

is there anyway to force the break time to less than 1 second? currently missing subsequent vehicles if there is a trigger.

 

[Vettester]

Yes, you can either type in a decimal #, such as 0.5, or use the arrows next to that box to change it.

Not sure exactly when this changed but for the last few months the minimum is now 1 second and you cannot set it to anything less.

 

[looney2ns]

Deleted incorrect information.

 

[Vettester]

You mean like this

You’re talking “make time” while we’re talking about “break time”.

 

[OfficerD00fy]

yea break time min is 1.0. What is the recommended make time for licence plates?

 

[Vettester]

What is the recommended make time for licence plates?

0.1 works well for me.

 

[sp3ctre]

I am having an issue with my alert to my API. I am trying to use &PLATE but it's not passing through anymore. I know the plate data is there as I get Alert_AI":"YA23XDR:98%" when I use "Alert_AI":"&MEMO"

Anyone know why &PLATE is no longer working?

 

[Vettester]

I am having an issue with my alert to my API. I am trying to use &PLATE but it's not passing through anymore. I know the plate data is there as I get Alert_AI":"YA23XDR:98%" when I use "Alert_AI":"&MEMO"

Anyone know why &PLATE is no longer working?

I'm using &Plate in an MQTT topic and it's still working for me using the following string:

{ "plate":"&PLATE", "AlertImgPath":"&ALERT_PATH", "Alert_AI":"&MEMO", "Date":"%Y-%m-%d %H:%M:%S","Camera":"&CAM"}

 

[kramttocs]

Ran into this on a fresh install:
AttributeError: 'ALPR_adapter' object has no attribute '_num_items_found'

Any ideas on this? Have rebooted but no impact. Still getting this error.