An Optimized YOLO-based License Plate Recognition System with Integrated Privacy Safeguards

Efficient and accurate license plate recognition (LPR) is crucial for intelligent transportation systems, surveillance, and security applications. This work proposes a Powered Smart License Plate Recognition System utilizing YOLOv8 for real-time detection and Optical Character Recognition (OCR) for character extraction. The system integrates publicly available datasets, including OpenALPR, CCPD, and UFPR-ALPR, along with custom-annotated data to enhance robustness. Data preprocessing involves normalization, augmentation, and privacy protection measures such as Gaussian blurring and encryption to ensure compliance with data protection regulations. The architecture is optimized using convolutional layers for feature extraction and attention mechanisms for improved character recognition. A privacy protection module ensures sensitive information masking, while Edge AI deployment on Jetson Nano or Raspberry Pi enables real-time inference with minimal latency. Model training employs transfer learning with pre-trained YOLOv8 weights, fine-tuned on license plate datasets. Optimizers such as Adam and SGD, along with IoU loss, enhance bounding box accuracy, ensuring precise detection. Performance evaluation demonstrates high mean Average Precision (mAP) exceeding 95% and OCR accuracy surpassing 90%, validating the system's efficiency across diverse environmental conditions. Applications include automated toll collection, traffic monitoring, vehicle tracking, and secure entry management in smart cities. Law enforcement agencies can utilize the system for stolen vehicle identification and traffic rule enforcement. The proposed approach offers a scalable, privacy-aware, and high-accuracy solution for modern LPR applications, ensuring compliance with security and regulatory standards.