DESIGN, CONSTRUCTION, MAINTENANCE
Print ISSN: 2944-912X, E-ISSN: 2732-9984 An Open Access International Journal of Engineering
Volume 5, 2025
Malaysian Car Plate Detection and Recognition Using Deep Learning Techniques
Authors: , , ,
Abstract: This project presents the development of an automated car plate detection and recognition system using deep learning, aimed at overcoming the limitations of manual recognition in traffic and security applications. A locally trained YOLOv8 model, integrated with EasyOCR, was employed to detect and extract characters from Malaysian car plates. The system was trained on a custom dataset with and without image preprocessing techniques, including grayscale conversion and contrast enhancement. Testing across single-line, double-line, and special-case formats showed that preprocessing significantly improved detection accuracy, increasing mAP50 from 89.0% to 99.5%. Among all formats, single-line plates recorded the highest F1 score (0.74) and similarity (0.81). Cross-platform performance analysis revealed that the GPU implementation achieved a much faster runtime (0.13 s/image) compared to CPU (1.23 s/image), while maintaining similar accuracy. These findings demonstrate that a segmentation-free deep learning approach, enhanced with preprocessing, delivers high accuracy and real-time performance for Malaysian car plate recognition under varied conditions.
Search Articles
Keywords: License Plate Recognition (LPR), YOLOv8, EasyOCR, Deep Learning, Image Preprocessing, Object Detection, Character Recognition, GPU Acceleration, Real-Time Systems, Map50
Pages: 106-111
DOI: 10.37394/232022.2025.5.12