A Deep Learning Based Approach for License Plate Recognition

الملخص

License plate recognition plays a critical role in modern smart-city applications, traffic monitoring, and law-enforcement systems. Traditional recognition systems often struggle with varying lighting conditions, image blur, occlusion, and differences in Saudi license plate formats. This research proposes an end-to-end deep-learning model for Saudi License Plate Recognition using YOLOv5 for plate detection and a YOLO-based OCR model for character recognition. The dataset was cleaned, verified, re-split, and processed using automated scripts to remove corrupted and blurry images, ensuring high-quality training samples. The YOLOv5 detector achieved strong performance with high precision, recall, and mAP scores across all experiments. Additionally, a custom mapping layer converted English OCR outputs into their equivalent Arabic plate letters, enabling fully bilingual recognition. The final integrated system successfully detects Saudi license plates and extracts the complete alphanumeric sequence with high accuracy. These results demonstrate that deep learning can outperform traditional OCR and rule-based systems, providing a scalable solution for real-world Saudi traffic environments