License Plate Recognition System Using Deep Learning Approach

The rapid growth of registered vehicles in Bangladesh has increased the demand for an accurate and automated Bengali License Plate Recognition (BLPR) system for intelligent traffic monitoring and law enforcement. Existing Optical Character Recognition (OCR)-based approaches often perform poorly on Bangla license plates due to complex layouts, font variations, and mixed character sets. To address these challenges, this paper proposes an OCR-free, end-toend BLPR framework based on the YOLOv8 deep learning architecture, enabling simultaneous license plate detection and character recognition within a single network. The model is trained on a custom dataset comprising more than 3,000 annotated images across 102 Bangla character, numeral, and district classes. Experimental results demonstrate strong performance, achieving 94.2% detection accuracy, 92.7% characterlevel recognition accuracy, and an average inference time of 7.9 ms per image. These results indicate that the proposed system is both accurate and computationally efficient, making it suitable for real-time deployment in intelligent transportation systems.