Automated number plate recognition system

Emina E. Etomi and Donatus. U. Onyishi
Keywords: Automatic number plate recognition, digital image processing, optical character recognition, openCV
Tropical Journal of Science and Technology 2021 2(1), 38-48. Published: July 25, 2021


Abstract

Automatic Number Plate Recognition (ANPR) system is an automated mass surveillance method that uses several Digital Image Processing (DIP) technique and Optical Character Recognition (OCR) on images to read and identify vehicle registration plates. ANPR has yielded multiple positive results in practical applications such as: access control, traffic law enforcement, inventory and property management, security systems surveillance, parking space allocation, and road traffic surveillance. The automatic number plate recognition system (ANPR) developed in this research work focused mainly on number plate localization and licence plate extraction from an image for possible application in different areas. It achieves this by using several OpenCV digital image processing (DIP) technique developed with python to bring about image segmentation from which some image segments were tested for characters, so that the length of character found on each segment with similar properties becomes the key towards localizing and cropping off the region with the actual vehicle licence plate. Some properties of characters that was used to isolate the possible licence plate are the fact that characters of the licence plate have corresponding image height, width, aspect ratio etc. using these pixel properties it was possible to filter off unwanted contour lines/curves that stands out as noise while localizing the actual region of the image having the plate number. Once the region was obtained OCR was used via a trained template for several character styles to obtain the text format of the licence plate. The work developed had a plate localization accuracy of 100% and 90% read accuracy.