215
2008
IMPROVING THE PERFORMANCE OF HYPERSPECTRAL PUSHBROOM IMAGING
SPECTROMETERS FOR SPECIFIC SCIENCE APPLICATIONS
F. Dell’Endice 3 ’ *
Remote Sensing Laboratories, Department of Geography
University of Zurich - Winterhurerstrasse, 190 - 8057 Zurich - Switzerland
francesco.dellendice@geo.uzh.ch
Commission WG VII/3
KEY WORDS: Hyperspectral, Spectral Binning, CCD, Pushbroom
ABSTRACT:
Hyperspectral imaging spectrometers offer the unique chance of recording image data of a broad range of targets in the reflected
solar energy spectrum. These instmments are designed upon certain requirements such as signal-to-noise ratio (SNR), spectral
resolution and bandwidth or noise equivalent delta radiance. These parameters are determined by investigating one or several typical
targets (e.g. vegetation, limnology, soil, atmosphere) that the instrument will sense during its operational life by means of specific
instrument models.
Depending on the specific application, users can demand hyperspectral image data that might cover a portion or the whole sensor
spectral range and, more importantly, may have requirements different from the ones the instrument was designed for originally.
Therefore, in order to meet the user requests the spectrometer settings should be modifiable.
Many instruments are potentially programmable from the electric point of view, in a way that the sensor setting parameters could be
changed, e.g. exposure time, on-chip averaging, the so-called binning, amplifier gains. By tuning these parameters the sensor
performances can be modified according to the user needs.
The Airborne Prism Experiment (APEX) 1 , a hyperspectral imaging spectrometer developed by a Swiss-Belgium consortium on
behalf of the European Space Agency (ESA) and under the scientific supervision of the Remote Sensing Laboratories (RSL), has
been designed upon certain requirements (e.g. radiance levels, SNR) but, nevertheless, the electric settings can be changed by means
of a mission control file in order to fulfil user requests that differ from the default scenario. Namely, the APEX instrument allows
changes of exposure time, on-chip binning and frame period.
We designed and implemented a software utility that optimizes the instrument parameters based on the possible range of hardware
settings and the user application requirements. This utility is based on the detector electrical and optical description, which is
modelled in terms of signal and noise by using the SNR equation 2 . In order to develop such a model the instrument optical
characteristics, i.e. transmission, must be known. The utility can be regarded as an APEX sensor simulator but it can be easily
adapted to any other hyperspectral imaging apparatus.
Users (i.e. sensor manufacturers, operators, scientists) can formalize their requirements and feed them into the model. E.g. a scientist
is aiming at estimating the amount of leaf chlorophyll content within a vegetation target with a required minimum of detectable
differences. Therefore he has to identify the needed values of SNR, spectral resolution and sampling interval as an input for the
simulator.
The utility evaluates all the possible solutions in terms of exposure time and on-chip binning in order to determine the one that
matches the scientist needs the best. A broad variety of error deviations are reported in order to help the users in interpreting the
simulation results, estimate the error and accuracy budgets accordingly.
Depending on the input requirements the discordance between the users needs and the results can be significant. In such a case the
utility performs a further step by analyzing post-processing strategies, as for instance off-chip binning, in a way that the requirements
can be someway be met.
The presented utility has a twofold advantage: (1) it allows manufacturers and sensor operators to offer an instrument that is
adaptable to needs of the end-users community and (2) it lets users, mainly scientists, understand what can be achieved with a given
hyperspectral instrument. The weakness of the utility relies on the lack of information about the optical and electrical parameters,
which might be caused by the confidential nature of technical details, namely in private companies.
We firmly believe that this utility can (a) optimize the programming of hyperspectral imaging spectrometers to gather more accurate
image data and (b) let users exploit the broad range of applications that can be investigated with the available large spectral range.
Corresponding author.