VEHICLE LICENSE PLATE RECOGNITION USING YOLOV8 AND PADDLEOCR

Abstract

This study focuses on developing a computer vision-based vehicle license plate recognition system designed to enhance operational efficiency and support the implementation of real-time intelligent transportation systems. Specifically, the research evaluates the system's performance in detecting and recognizing license plate characters using a two-stage approach: (1) an object detection stage employing the YOLOv8 model to identify plate locations, and (2) a character recognition stage utilizing PaddleOCR for text extraction. Experimental results demonstrate exceptional detection accuracy, with 96.3% precision and 98.7% recall, indicating the system's robustness under varying environmental conditions such as lighting changes and camera angles. However, character recognition accuracy remains relatively low (39.62%), potentially due to input image quality limitations or license plate complexity. These findings highlight the critical need for optimized image pre-processing techniques (including noise reduction, contrast enhancement, and perspective correction) to improve text readability prior to OCR. Future research will explore various image processing methods and alternative model architectures to enhance recognition accuracy and real-time performance stability, enabling effective integration into intelligent transportation applications such as traffic monitoring, automated parking systems, and AI-based law enforcement.