COVID-19 Thales Thermography Triage : Thermal Camera Feasibility Study

Justification for research and project plan Background In the fifty years since the emergence
of thermal imaging technology, Thales in Glasgow has built up a world leading capability in
the design, manufacture and supply of Thermal Imaging cameras. In addition to the cameras,
Thales in Glasgow has a particular expertise in developing image processing algorithms
(conventional and artificial intelligence based) to allow the cameras to perform critical
user tasks beyond mere imaging.

COVID-19 [SARS-COV-2] has placed a huge challenge on the world. In response to the crisis,
Thales is engaged in a number of initiatives, including one aimed at the possible application
of thermal imaging cameras to detect people with a fever and hence those who may be suffering
from COVID-19. This initiative has received encouraging feedback from prospective users
around the world who are looking, not only at the immediate issue in the hospitals, but also
forward to a time beyond the current lockdowns, when cost effective techniques for surveying
groups of people for potential COVID-19 sufferers will be required. This surveillance, for
example, could be in hospitals, at airport gates, in buildings or outside in streets. As part
of the initiative Thales Glasgow has been performing analysis and experiments using cameras
looking at faces to confirm what temperature differences can be measured.

What Thales lacks is real world thermal images of patients suffering fever and access to
clinicians who can advise on the medical aspects of the work. This research would fill these
two voids.

Benefit and Rationale Existing thermal camera based fever detection systems suffer from two
main issues that restrict their current application: Cost and Sensitivity.In order to achieve
the required sensitivity, existing systems often require Black Body (BB) calibration sources
to be visible in the scene. These BB sources are often more expensive than the cameras (circa
one to two orders of magnitude more expensive). Thales believe they can achieve the required
sensitivity without the need for BB sources by using advanced thermal camera correction
algorithms.

The variability in skin emissivity makes conversions from irradiance as measured by a thermal
camera, to absolute skin temperature inaccurate. Previous work place the emissivity of human
skin between 0.990 and 0.999, however earlier research suggests that the value is 0.971 ±
0.005(SD). Typically in existing systems, this inaccurate, absolute temperature is used along
with a simple threshold algorithm to find skin above a certain temperature leading to both
false positives and false negatives.

Significant statistical correlation between the temperature of multiple facial features and
measured human temperature has been observed. Their research found that for most facial
features there was a correlation between multiple aspects of the facial view and recorded
oral and tympanic temperature, with correlations of 0.5 or greater observed for multiple
aspects of both the front and side of the face, with the ear offering the greatest
statistical correlation.