Enhanced Human Identification in the Thermal Infrared Using Mid-IR and Longwave-IR Polarimetric Imaging
Report Number:
ARL-TR-9310
September 23, 2021
Approved for public release: distribution is unlimited.
Author(s):
Kristan P Gurton
Abstract:Until very recently, facial recognition techniques relied on data sets that were predominately derived using visible imagery that often required a degree of ambient illumination while being recorded. However, there are applications in which human identification is required in environments devoid of any form of visible light or illumination (e.g., nighttime observation). Under such circumstances, IR thermal imaging is usually applied for target location and identification, but was considered insufficient for human identification applications since conventional thermal imagery lacks the detailed spatial information required for face recognition. Reported here are results of ongoing research that involves a new form of thermal imaging that takes advantage of additional polarization-state information inherent in most image forming radiation (i.e., polarimetric thermal imaging). By creating thermal image sets based solely on polarization-state information, we show significant improvement in the spatial detail displayed in the resultant thermal imagery for both mid-IR and longwave IR spectral wavebands.
