Egyptian Car Plate Detection and Recognition Using Computer Vision

Abstract

Automated license plate detection and recognition is a demanding task because it plays a pivotal role in activities such as security zone access control and intelligent traffic management and transportation systems. Traditional plate recognition approaches using image processing have limitations in handling real-life situations such as varying lighting conditions, moving vehicles, and background noise. Computer vision using Deep learning can effectively tackle these limitations. Several research efforts have employed deep learning frameworks for license plate detection, with limited focus on recognition, particularly for plates supporting Arabic letters. To address this gap, this paper introduces deep learning approaches for car license plate detection and recognition in the Egyptian plate system. Specifically, wet rained various computer vision deep learning models and compared their performance on a proposed dataset that was manually collected and annotated. Experimental results on YOLO (versions 5 and 8) and DETR demonstrate that YOLOv5 outperforms the other models, achieving a mean average precision (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$m$</tex> AP) of 99.45% for detection and 97.43% for recognition,