Abstract:
Automatic monitoring of preventive health measures in crowded public environments presents a significant challenge, particularly regarding mask compliance and social distancing enforcement. This study introduces an intelligent real-time monitoring system for face mask detection and social distance estimation utilizing the YOLOv8 object detection framework. The system is developed to deliver accurate and efficient automated surveillance through computer vision techniques. A custom dataset of 12,603 annotated images, categorized into mask, no mask, and incorrectly worn mask classes, was employed for training and evaluation. Data preprocessing and augmentation were implemented to enhance model generalization, and hyperparameter optimization was conducted to improve detection performance. Additionally, a homography-based distance estimation module was incorporated to measure interpersonal distances and identify social distancing violations in real time. Experimental results show that the system achieves a mean Average Precision (mAP@0.5) of 94.1%, with particularly high detection accuracy for the incorrectly worn mask class (99.2%). The model maintains a strong balance between precision and recall, achieving a maximum F1-score of 0.92. The integrated distance monitoring module also provides reliable proximity alerts for potential violations. These findings suggest that the proposed framework offers an effective and practical solution for intelligent public health monitoring and is suitable for deployment in environments such as airports, hospitals, transportation hubs, and other crowded public spaces.
