IEEE Geoscience and Remote Sensing Magazine - June 2019 - 81

has been made on the SWaP requirements but it remains
an area of active research. The interferometric designs
have a clear advantage in this regard. The second area is
in FPA development, with the goal of very low nonuniformity/high operability, minimal noise, and relatively high
operating temperatures.
SENSOR CHARACTERIZATION
The signal from an imaging spectrometer is recorded in
digital numbers (DNs) by the readout electronics from
the FPA. This directly recorded imagery is highly useful
in informing our knowledge of the action of the sensor
but is of very limited utility in terms of an exploitable
data source. The conversion of the raw DNs to physical
units is through
◗ sensor characterization, where the response of the system to a known input is measured and used to quantitatively determine the spectral, radiometric, and spatial
properties of the imaging spectrometer
◗ data calibration, where the raw data are converted to
radiance values and each spectral sample is assigned
a wavelength.
The final calibration products increase the utility of the
data by enabling the application of a broad range of physics-based processing algorithms. We will not address the
important process of spatial characterization and georeferencing where the line-of-sight direction for each spatial
sample relative to the platform reference system is established to support accurate image geolocation.
SPECTRAL CHARACTERIZATION
Spectral characterization involves determining the position and shape of each instrumental profile P^m c -mh. In a
dispersive system, a scanning quasi-monochromatic source
is used for each spatial-spectral sample. This can be accomplished using a tunable laser source or a scanning monochromator. The simplest approach is to use an integrating sphere that uniformly illuminates the entrance pupil,
which requires a sufficiently bright source. Otherwise, the
monochromator slit is imaged onto the spectrometer slit
with the appropriate relay optics to match the slit width and
F-number, which may be difficult, especially in the case of
an imaging spectrometer with a curved slit. Another solution is for the monochromator and spectrometer slits to be
crossed and to measure the instrumental profiles at a few
field angles and then model the unmeasured ones. This
works well for high-performing imaging spectrometers. A
similar approach is used to characterize a Fabry-Pérot interferometer, with the added complication of verifying the
free spectral range.
The spectral characterization is simpler for a Michelson
interferometer-based imaging spectrometer because the instrumental profile is determined by the optical path difference that is precisely sampled by the monochromatic metrology laser. The laser wavelength is accurately known, and
the sampling, typically at the zero crossing of the sinusoidal
JUNE 2019

IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

metrology fringe, determines the interferogram optical
path difference sampling positions. The more challenging
problem is the geometric characterization of the detector
elements in relation to the optical axis. Each element has
an associated optical path difference as well as self-apodization because of its finite
size. The measurement of a
monochromatic source is used
THE NOISE
to characterize the spectral
CHARACTERISTICS OF
performance for the off-axis
FPAS ARE CRITICAL TO THE
elements in relation to an onaxis element to accurately asOVERALL PERFORMANCE
sign the wavelength scale to
OF THE IMAGING
the 2D array [22].
SPECTROMETER DRIVING
The spectral calibration is
THE COOLING
verified and refined by meaREQUIREMENTS.
suring the transmission of a
known filter or gas cell. An
example of the spectral filter is the National Institute of Standards and Technology
Standard Reference Material 1921a, which is a polystyrene
film with distinct spectral features through the 500 to
3,500-cm−1 (2.8-20-nm) range. A gas cell filled with carbon dioxide is another choice covering 12.5 to 15 nm. Field
measurements are also utilized to further refine the spectral characterization using atmospheric absorption features
fit to a radiative transfer model. For a well-behaved imaging
spectrometer, the final spectral calibration is captured in
an array, with the instrumental profile centers and widths
recorded for each detector element. Errors in the spectral
calibration of the data impact the utilization of TES techniques [23].
RADIOMETRIC CHARACTERIZATION
The radiometric characterization of an imaging spectrometer produces the data used to convert raw signals to radiance
units; these are utilized in the application of physics-based
algorithms, which are typically employed in the first step of
exploitation. We will describe the process for the characterization of an airborne sensor with a linear response where
the entrance pupil is fully illuminated by known sources.
The characterization yields a two-point data calibration utilizing blackbody sources, where the emissivity is very close
to unity for all wavelengths of interest, which span the expected temperature range of a scene.
The measurement equation is modified to
S (m c) . Ux o ht int

#-33 L (m) P (m c - m) dm / R (m c) L (m c),

(6)

where it is assumed that the efficiencies can be approximated by their values at the channel center m c, R (m c) is the system responsivity for the channel, and L (m c) is the in-band
radiance. The signal has contributions from both the at-aperture radiance and self-emission from the optics given by
S (m c) = R (m c) L (m c) + S is (m c),

(7)
81

IEEE Geoscience and Remote Sensing Magazine - June 2019

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - June 2019