License Plate Recognition

This project implements a license plate recognition system using traditional image processing techniques, without relying on OCR libraries or machine learning models. The system works under the following conditions:

• The plate occupies at least one-third of the image width.
• The angle between the camera’s optical axis and the plane of the plate is no more than 45 degrees.
• Recognizes standard plates with black characters on a white background (7 or 8 characters).
• Plates from various regions are supported.
• Images can have varying resolutions.

Example images:

Results:

{
    "459011288_1970877216659435_8483961446294724982_n.jpg": "PTUJE07",
    "459315184_1071610051246153_2361377541037099992_n.jpg": "PTU8166C",
    "459451150_2302604903407199_7358595598509359465_n.jpg": "PTU4806F"
}

Predict plate numbers:

python3 predict_plate_main.py /path/to/images name/for/results/file.json

Calculate accuracy:

python3 checkAccuracy.py /path/to/correct/results.json /path/to/results.json

Example dataset:

Dataset link