Combining YOLOv8, Real-ESRGAN, and EasyOCR for Robust License Plate Recognition in Real-World Conditions

Abstract In an era of rising vehicular density and heightened environmental awareness, the need for automated, intelligent vehicle monitoring systems has become increasingly critical. This paper proposes a robust, scalable, and end-to-end license plate recognition and validation pipeline tailored to real-world scenarios, particularly optimized for the diverse formats and conditions found in Indian vehicular systems. The system integrates cutting-edge components: YOLOv8 for high-precision license plate detection, Real-ESRGAN for super-resolution enhancement of low-quality images, EasyOCR for accurate optical character recognition, and SQLite for real-time database validation and emission compliance checks. Unlike prior works that focus on isolated components, this study offers a unified framework that harmonizes detection, enhancement, recognition, and verification into a cohesive pipeline. The proposed solution demonstrates strong performance under varying lighting, resolution, and plate design conditions. Experimental evaluations validate its efficacy in terms of accuracy, speed, and adaptability, highlighting its suitability for deployment in intelligent transportation applications such as traffic surveillance, automated tolling, law enforcement, and environmental monitoring.