Vehicle License Plate Number Detection with YOLO11

Abstract

Automatic License Plate Recognition (ALPR) systems are crucial for automating vehicle monitoring and access control, especially in parking management. However, many existing ALPR methods rely on earlier detection models and perform poorly under low-light or night-time conditions. This study addresses that issue by developing a license plate detection and recognition system using the You Only Look Once (YOLO11) algorithm and PaddleOCR (Optical Character Recognition) for character recognition. The objective is to evaluate the performance of this approach on Indonesian two-wheeled vehicles under varied lighting conditions. The dataset includes daylight and low-light images, divided into training, validation, and test sets. YOLO11 was trained with several epoch configurations, and the best model achieved 99.8% precision, 100% recall, 99.5% mAP@50, and 80.7% mAP@50–95. The model was then integrated with PaddleOCR to extract license plate text. Results show this combination is effective for real-time vehicle identification and can improve parking queue management.