Non-contact breathing monitoring by integrating RGB and thermal imaging via RGB-thermal image registration

Abstract

Altered breathing rate is an important sign of disease status. Currently used machine-based monitoring of the breathing rate includes contact with the body, which may result in pain and discomfort. In this paper, a non-contact breathing monitoring technique is proposed by integrating RGB and thermal imaging systems with RGB-thermal image registration. This method provides a linear mapping for automated selection of the region of interest (ROI) followed by tracking to extract the breathing rate. To evaluate the efficacy of the proposed approach and its robustness against motion, talking, varying breathing rate or rhythm, and high ambient temperature, this study was conducted in three phases. Validation of the proposed approach demonstrated a strong agreement with the reference method of breathing rate monitoring using a respiration belt. During normal breathing, the mean absolute error (MAE) reached 0.11 bpm (breaths per minute). While in more challenging conditions, defined by three phases, the MAE reached 1.46, 2.08, and 1.69 bpm, respectively. In short, the proposed method performance is a promising alternative to a contact-based method due to its strong agreement and might be useful in diverse applications such as sport studies, rehabilitation centres, quarantine centres, and in hospital or airport screening during the COVID 19 pandemic.