Authors:
M. A. Sayedelahl

Addresses:
1Department of Computer Science, Information and Technology, Damanhur University, Beheira, Egypt. mohamed.abdelatif@cis.dmu.edu.eg1 

Abstract:

This paper presents a novel two-stage framework for Egyptian Vehicle License Plate Recognition (EVLPR), achieving a high recognition accuracy of 99.3% on a diverse dataset. The first stage utilizes image processing techniques for robust license plate localization, employing edge detection and morphological operations to isolate candidate regions within an image. The second stage leverages a custom-designed deep learning model specifically trained for Arabic Character Recognition (ACR). This model is trained on a dataset encompassing the variations encountered in real-world Egyptian license plates, leading to superior performance compared to existing approaches. By effectively addressing the complexities of Arabic script recognition, the proposed system paves the way for practical EVLPR applications with significant security and traffic management implications. The framework’s potential extends beyond license plate recognition, offering promising functionalities in smart traffic management systems for tasks such as identifying traffic violations and monitoring vehicle occupancy for optimized parking management. Future work will refine the system by exploring alternative architectures, expanding the dataset for broader applicability, and addressing system dependencies. Overall, this research presents a significant contribution to the field of EVLPR, demonstrating the effectiveness of a combined approach using image processing and deep learning for real-world challenges.

Keywords: Neural Networks; Variation Tolerance; Violations and Monitoring Vehicle; Robust Recognition; Arabic License Plate Recognition; Optimized Parking Management; Deep Learning.

Received on: 05/09/2023, Revised on: 19/11/2023, Accepted on: 09/12/2023, Published on: 05/03/2024

DOI: 10.69888/FTSIN.2024.000156

FMDB Transactions on Sustainable Intelligent Networks, 2024 Vol. 1 No. 1, Pages: 40-55

• Views : 289
• Downloads : 18

Download PDF