Nasal cavity detection in facial thermal image for non-contact measurement of breathing

Abstract

Abstract-In this paper, we propose a method of improving the accuracy of detecting nasal cavity location in far infrared images for non-contact measurement of human breathing. We found that although our previous method for far infrared imaging can detect regions that include nasal cavities well, it suffers from high false alarm rates. In order to reduce this rate, we extend our method with a false alarm classification function. Object detection based on a boosted cascade of Haar-like feature classifiers is applied to find the candidates of regions that include the nasal cavities. In false alarm classification, binarization is employed to strictly segment facial area and background. Based on the results of binarization, false alarms on the background can be accurately classified. 5,100 FIR images are collected to train our nasal cavity detector; we evaluate the number of false alarms and detection failures. The results show that the proposed method can reduce false alarm events.