Car Plates Recognizer

Repository contains description of Computer Vision laboratory project which is based on recognizing car plates using OpenCV with Python 3.7.

Project tasks

• find and extract possible car plates from image
• recognize signs from extracted plate

Example use case

• car back image

• extracted plate

• JSON file with result for given image

{
    "car_back.jpg": "PWL3787"
}

Project requirements

• car plates rotated less than 45 degrees in comparison to horizontal position
• width of plate is bigger than 1/3 of image width
• only common polish car plates which contain 7 black signs on white background
• image with different resolution
• no use of OCR and ML trained methods and algorithms

Detector results on project image set

• Accurancy: 0.845433 (361 good chars per 427 total)
• Find acc: 0.950000 (57 readed plates for 60 total)
• OCR acc: 0.904762 (361 readed chars for 399 total)