Fast Licence Plate Recognition of Moving Vehicles Using Deep Learning Techniques

Automatic Number Plate Recognition (ANPR) systems are crucial for autonomously identifying vehicle license plates in real-time. This paper presents a robust ANPR system designed to capture and analyze images of moving vehicles, recognize license plate numbers, and authorize access based on successful identification. The proposed system includes a two-stage deep learning framework, combining DarkNet53 and YOLOv8 to enhance accuracy and efficiency. DarkNet53 is employed to accurately detect UAE license plates, while YOLOv8 extracts the alphanumeric characters. A major contribution of this work is the comprehensive evaluation of system performance across varying environmental conditions, demonstrating the reliability and adaptability of the ANPR system. Additionally, a curated dataset of front-view and back-view UAE vehicles captured under different lighting and orientation conditions has been developed, further enhancing the contributions of this work by providing valuable data for testing and improving system accuracy. Experimental results show that the combined deep learning models achieve a 95% accuracy rate in recognizing license plate characters under diverse lighting conditions, highlighting the system’s effectiveness for real-world applications.