Automatic detection of face mask wearing based on polarization imaging

Area of Science:

• Computer Vision
• Image Processing
• Public Health Technology

Background:

• The global rise in respiratory illnesses necessitates effective automatic face mask detection systems.
• Developing accurate systems is challenging due to varying mask appearances and environmental factors.

Purpose of the Study:

• To introduce DOLP-YOLOv5, a novel framework for automatic face mask detection using polarization imaging.
• To enhance the accuracy and efficiency of face mask detection in diverse conditions.

Main Methods:

• Integration of a lightweight shuffle attention (SA) mechanism for feature extraction.
• Application of Content-Aware Bidirectional Feature Pyramid Network (CA-BiFPN) for robust feature fusion.
• Utilization of Focal-EIoU loss for precise bounding box regression.

Main Results:

• DOLP-YOLOv5 achieved a mAP50-95 of 63.5% on the self-constructed PFM dataset.
• Demonstrated improvements of 3.08% and 4.44% over YOLOv8s and YOLOv9s, respectively.
• Attained a high response speed of 384.6 frames per second.

Conclusions:

• DOLP-YOLOv5 exhibits superior performance in face mask wearing detection.
• Polarization imaging offers a promising approach for enhancing object detection tasks.
• The framework has potential applications beyond face mask detection, particularly in polarization imaging analysis.