Infrared (IR) thermography is a
well established technique for remotely measuring the temperature of a
surface where it is impractical or impossible to do so by a contact
means. The term thermography denotes an imaging capability, but
the concepts are the same for non-imaging sensors. IR
thermography exploits the correlation between the temperature of a
surface and the IR energy emitted by the surface.
|
This relationship is described by Stefan’s Law:
| |
|
where σ is the Stephan-Boltzman constant (= 5.67×10-8 W/(m2∙K4)) and T is the temperature of the surface.
|
The spectrum of the IR light is described by Planck’s blackbody function:
| |
IR thermographic measurements are
typically done in one of two spectral bands which exhibit relatively
low absorption of light by water vapor and carbon dioxide (CO2); these
are the 3 to 5μm band and the 8 to 12 μ m band and are sometimes called
“atmospheric windows”. For this particular measurement where we
are interested in measuring human skin temperature over a range of 35
to 55ºC, it makes radiometric sense to work in the 8 to 12μm band, as this is where the radiation is concentrated according the second equation.
In general, this measurement works well as long as all of the radiation
from the surface makes it to the detector. However, as we step
outside and attempt to make this measurement over a large standoff
between surface and sensor, the atmosphere has a strong detrimental
impact on the accuracy. The key to making an accurate and
absolute IR thermographic temperature measurement is being able to
correct for the effects of absorption due to water vapor and CO2 as
well as scattering and absorption due to fog, rain, and snow.
Thermographers which simply correlate emitted radiance with surface
temperature will report significant temperature errors in the presence
of high humidity, rain, and fog, as a large percent of the light is
attenuated before it reaches the sensor as shown in Figure 1.
Figure 2 illustrates the associated temperature error of the integrated
radiance approach over a 100 m path with humidity, CO2, and fog.
|
| Figure 1 shows the transmission
spectra of water vapor, CO2, radiation fog, and rain for high humidity,
moderate radiation fog, and light rain over a pathlength of 100
m. We impose the transmission spectra over the Planck blackbody
profile associated with a 40ºC surface (given by equation 2 and read
off the right y-axis). The attenuation by these atmospheric
constituents results in substantially less energy arriving at the
sensor relative to what left the surface. The effect is a
significant perceived temperature error. Molecular absorption
spectra was obtained through Hitran PC; fog and rain attenuation
spectra was obtained using Modtran PC.
|
Figure 2 shows the projected error
of the integrated radiance measurement over a pathlength of 100
m. Once the sensor is calibrated in the laboratory, there is no
easy way for it to compensate for atmospheric absorbance and scattering
in the path of the measurement when it is taken outside. The
result, as shown on the right axis, is a significant temperature
error. This error is particularly problematic in the DE
application because the target is always reported colder than it really
is; the microwave source operator may continue to heat the target to a
desired (but erroneously read) temperature level, resulting in
potentially significant injury.
| |
In response to this opportunity,
OPTRA has demonstrated the feasibility of a novel approach to IR
thermography. We have taken advantage of the quasi-symmetric
structure of water vapor absorption and radiation fog attenuation
(Figure 1) by spectrally splitting the IR image onto two miniature
uncooled microbolometer cameras using interference filters. The
two spectral “channels” approximately balance out the effects as shown
in figure 3. In addition, we filter out the absorption due to CO2
by positioning the cutoff of the longwave filter just short of the 13.4
m resonance band. By algebraically combining the images on a
per-pixel basis and normalizing the result, we are able to project a
surface temperature measurement in high humidity, fog, and rain with
minimal error (Figure 4).
|
| Figure 3 shows the two effective
spectral filters imposed on the atmospheric constituent
transmission. By carefully selecting the edges of the filters, we
balance out the attenuation in the two channels, regardless of the
pathlength or humidity or fog level. Normalizing the
algebraically combined images corrects for spectrally flat attenuation
such as that caused by rain, snow, and advection fog as well as other
sources of bulk attenuation.
|
Figure 4 shows the projected
temperature error for OPTRA’s thermographer measuring skin temperature
over standoffs of 100, 400, and 700 m. This figure illustrates
that by spectrally balancing the attenuation, we are able to report the
absolute temperature of a surface with minimal error at a standoff of
up to 700m.
| |
|
