System for Medical Mask Detection in the Operating Room Through Facial Attributes

This paper introduces a system that detects the presence or absence of the mandatory medical mask in the operating room. The overall objective is to have as few false positive face detections as possible without losing mask detections in order to trigger alarms only for healthcare personnel who do not wear the surgical mask. The medical mask detection is performed with two face detectors; one of them for the face itself, and the other one for the medical mask. Both detectors run color processing in order to enhance the true positives to false positives ratio. The proposed system renders a recall above 95 % with a false positive rate below 5 % for the detection of faces and surgical masks. The system provides real-time image processing, reaching 10 fps on VGA resolution when processing the whole image. The Mixture of Gaussians technique for background subtraction increases the performance up to 20 fps on VGA images. VGA resolution allows for face or mask detection up to 5 m from the camera.

keywords: Face detection, facial attributes, AdaBoost