A Biometric Application for Thermal Facial Signatures and Templates Using Mid-Wave Infrared Images

This abstract presents biometric findings using thermal infrared imaging for the detection of the human facial vasculature network at the skin surface. This involves the generation of thermal facial signatures and templates. A thermal infrared camera with reasonable sensitivity provides the ability to image superficial blood vessels on the human skin. The experiment consists of image processing techniques used on thermal infrared images captured using the Merlin™ InSb MWIR camera from FLIR systems with a thermal sensitivity of 0.025°C at 30°C ambient temperature. For the purpose of this experiment thermal images were obtained from 13 volunteers, they were asked to sit straight in front of the camera and a snapshot was taken of their frontal view. This process was repeated at least three more times during different days and times of the day to take into consideration subtle variations that may occur. The thermal infrared images were then processed to extract facial vasculature features for each volunteer.  The integrated process consisted of three main steps: image registration, vasculature extraction, and generation of thermal facial signatures and templates. A similarity measure was then used to match the facial signatures. The average matching accuracy rate is 88.46%. This process was then repeated on an independent database of thermal infrared imaging and similarly good accuracies were obtained. Our results show that thermal infrared images allow us to extract facial vasculature features through a set of integrated image processing techniques and that our similarity measure is a viable matching technique in biometric applications.