Skip to content

Latest commit

 

History

History
30 lines (24 loc) · 1.43 KB

README.md

File metadata and controls

30 lines (24 loc) · 1.43 KB

Vietnamese ALPR_System

An Automatic License Plate Recognition System for Vietnamese Plates

This system can detect and recognize 2 types of license plate in Vietnam, rectangle plates and square plates.

Abstract

  • This system can detect and recognize license plates from images, videos and webcams.
  • Within this project, the camera's position is fixed and only one car at a time can drive through the gate. Therefore, the system is only able to detect 1 plate per frame.

Method

  1. Plate detection
    • Sobel X for detecting vertical edges followed by a morphological transformation.
    • Finding contours which satisfy the ratio of the plate to get the possible plates
    • Checking for characters on the possible plates found to assure it is a license plate.
  2. Plate recognition
    • For character recognition, I used MobileNet_v1_0.5_128 as it was lightweight and suitable for real-time recognition.

Requirements

  • python 3.6
  • run pip install -r requirements.txt

Implementation

  • run test.py for testing on a video.
  • run test_image.py for testing on an image.

Result

  • Demo
  • Demo2

Note

  • You should play with these parameters: minPlateArea, maxPlateArea and ksize in cv2.getStructuringElement to implement successfully on your own case.