Logo Lanfrica

Azed1ne/Enhanced-ANPR-for-Algerian-plates

Domaine:

digital infrastructure

Type de record:

software
Créateur:
Aze
Hôte:
Algerian vehicle number plate recognition with data extraction using Python's openCV and easyOCR # Enhanced ANPR for Algerian plates Algerian automatic vehicle number plate recognition with data extraction using Python's openCV and easyOCR without any Machine Learning, with a custom algorithm that enhances and corrects the values detected for low-quality images. The **main idea** is to find the plate, crop it, zoom it and apply easyOCR methods ( reader.recognize() & reader.readtext() ) to read the digits and then send the 2 results to the algorithm to process them, fix common errors and output the most accurate plate number to finally extract information about the car. ## Examples: ## Algerian registration plates facts: + Being composed solely of numbers, they are one of the few vehicle registration plates which can accurately be referred to as '**number plates**'. + The registration mark takes the form of three groups of digits separated by a space. - The **first group** of numbers comprises 5 or 6 digits that make up the vehicle's actual registration/serial number. * The **second group** of numbers is always composed of 3 digits, the first one indicating the class of the vehicle in accordance with the following list: 1. Private Vehicles. 2. Trucks. 3. Vans (Camionettes). 4. Buses. 5. Tractor units. 6. Other Tractors. 7. Special vehicles. 8. Trailers. 9. Motorcycles. - The other 2 digits represent the year in which the vehicle was manufactured, for ex: 99 = 1999, 06 = 2006 ..etc. - If the second ( middle ) group contains "00" it means the car is newly purchased and has a temporary plate. - The **third group** of numbers contains 2 digits representing the wilaya (city) or province where the vehicle was first registered ( from 01 to 58 ). - **Example:** More info can be found here or here. ## Requirements and limitations: - The image needs to be captured with a decent quality ( since we'll be zooming on the plate only ). - It's made to detect only 1 plate in a picture. - The plate needs to be a rectangle ( the logic for custom plates with two lines is …