Principal Author: Tom Flatley
Title of Paper: On-board Cloud Contamination Detection with
Atmospheric Correction
Abstract: The
desirability of performing satellite, on-board science data processing and analysis,
and the actual production of science data products in space is unquestionably a
solution to the often discussed "data glut" problem but the roadmap
to get there is rather murkey. An investigation of 2 typical preprocessing
algorithms, cloud masking and atmospheric
correction, routinely performed during satellite data postprocessing in
the Earth Sciences domain is intended to provide some insight on more sophisticated
analyses. Central to the issue of realtime data processing in an on-board
computing environment is speed, therefore, these 2 algorithms are benchmarked
on a variety of commercial microprocessors with the expectation of inferring
performance on flight-qualified processors. Intuitively, a sequential
processor, especially the lower
performance radiation hardened versions essential to some space missions,
cannot process and analyze science data fast enough to be practical in a
realtime, on-board environment so 2 other computing devices, the Field
Programmamable Gate Array (FPGA) and the
Application Specific Integrated Circuit (ASIC), which introduce some
degree of parallelism, are also investigated with hardware implementations of
the 2 algorithms. The embedding of these
algorithms in silicon introduces additional parameters other than speed
which must also be considered when
approaching the issue of sophisticated on board science processing. Gate
count becomes equally important as
speed since radiation immune FPGAs and perhaps to a lesser extent ASIC's are
limited in design space. There is also the issue of acceptable accuracy since
in situ data, conventionally required for some preprocesssing algorithms is
unavailable.