Real-time leak localization in N95 respirators using infrared imaging and deep learning with optimal ROI signal correlation

ORCID

Clothilde Brochot : 0000-0002-2431-5053

Ali Bahloul : 0000-0002-4597-2001

Type de document

Articles dans des actes de congrès

Année

2025

Langue

Anglais

Directeurs de la publication

The Institute of Electrical and Electronics Engineers

Titre des actes

2025 6th International Conference in Electronic Engineering & Information Technology (EEITE)

Maison d’édition

IEEE

Résumé

A secure seal in N95 respirators is critical for preventing airborne contaminants, yet minor leaks significantly compromise protective efficiency. This paper presents an integrated framework leveraging infrared imaging and deep learning for real-time, non-contact leak localization. A custom U-Net model extracts the mask region from thermal images, while the Segment Anything Model 2 (SAM2) dynamically tracks contour variations under changing conditions. Thermal signals along the mask boundary are transformed into the frequency domain and correlated with a reference breathing signal to identify leak locations. The proposed method systematically evaluates optimal central region selection to enhance correlation-based leak detection. Experimental validation using pixel-wise Breathing Cycle Optical Flow Tracking demonstrates the effectiveness of this approach in accurately detecting and localizing leaks in real-world conditions, offering a robust alternative to conventional contact-based fit-testing methods.

Mots-clés

Équipement de protection respiratoire, Respirator, Masque N95, N95 mask

Numéro de projet IRSST

2022-0008

Ce document n'est pas disponible pour le moment.

Partager

COinS