Recognition of vehicle number plates in large-size photographs

[garry0garry]

This is my first time solving the problem of detecting vehicle registration plates in "CodeProject.AI". I am a beginner.

I installed the modules "Object Detection (YOLOv5 .NET)", "Object Detection (YOLOv5 6.2)", "License Plate Reader".
In my case, vehicle registration plates are not detected in a 3840x2160 photo.

But if I select a fragment with a vehicle in such a photo and place it in a separate file without changing the fragment size

, then the vehicle registration plates are detected correctly (confidence 85%).

I assumed that I needed to enable Model Size Tiny or Large in "Object Detection (YOLOv5 .NET)" for large photo. Switching modes did not help.

Is it possible to somehow change the module settings so that they would detect vehicle registration plates on a 3840 x 2160 photo?

[tbc725]

I don't think it matter on the model size, because they are using Custom Model called "License Plates". So what matters is the License Plates models' performance. However, I think the real problem is the relative size of the plate in the image. It is too small.

It is hard for the License Plate model to capture the plate with such size. You either need to zoom in the camera more, or adjust the workflow of your LPR. (e.g. Use Object Detection to detects the car, pass on the car image to the License Plate model to detect the plate and then run the ALPR for plate recognition.)

CPAI's LPR + Object Detection is:
(1) Object Detection runs the License Plate to detect the license plate (it was trained to detect the shape of the license plate I believe).
(2) Once the plate is found, it crops it and pass on the license plate image to ALPR for the recognition process.

What you are running into is the plate is not being detected due to its size is too small in the overall image.

What might help you solve the problem is :
(1) Object Detection detects car or truck and crops the image of it.
(2) Object Detection take the cropped image and runs the License Plate model to detect the license plate (just the plate, not the text).
(3) Object Detection then take the cropped license plate image and pass it to the ALPR for the recognition process.

Let's be practical:

1. Feed it with overall image, nothing detected:
   

2. Now, I cropped the image of the car and then feed it again (simulate the process of passing on the detected car image)
   

3. I give the cropped image to ALPR:
   

I hope that help.

Note: I noticed the License Plate model in YOLOv8 is better on capture the license plate (the plate only), more effective than YOLOv5.

[MikeLud]

What you can do in Blue Iris create a clone of this camera and use the Area of interest (AOI) to create an AOI similare to the below image. Then use the clone camera for ALPR.

[tbc725]

That is actually a brilliant idea to solve this.