IEEE Geoscience and Remote Sensing Magazine - June 2023 - 26

SENDING LARGE AMOUNTS
OF RAW HYPERSPECTRAL
DATA FOR THE ON-THEGROUND
PROCESSING IS
INEFFICIENT AND EVEN
INFEASIBLE, DEPENDING ON
THE LINK CONSTRAINTS.
MISSION CONSTRAINTS
The CHIME payload will be equipped with a hyperspectral
camera delivering more than 250 spectral bands at a
# 10-mn spectral sampling interval in the visible and infrared
spectral range from 400 to 2,500 nm. The instrument
field of view covers a swath of 130 km at a 30-m GSD at the
CHIME spacecraft altitude. The revisit time will be 11 days
when the two satellites are operational [167]. With a continuous
data throughput close to 5 Gb/s acquired over land and
coastal areas, the CHIME payload will deliver about 1 Tb
of data per orbit. Downloads are foreseen within 6-12-min
visibility time slots over the Svalbard and Inuvik polar stations
through the Ka-band link, which offers up to a 3.6-Gb/
s download data rate. The acquired data will be processed
on the ground between 5 and
10 h after acquisition. As a
complement, onboard data
processing might be an asset
to reduce and select/prioritize
valuable information. This
leads to envisioning new strategies
for onboard processing,
storage, and transmission for
CHIME. Besides the common
constraints imposed on any
hardware that will operate in
space (vacuum, temperature,
radiation due to the space environment,
and vibrations during the launch phase), one of
the main challenges is to consider the real-time processing
constraints imposed by the data acquisition principle used
onboard CHIME alongside the constant monitoring mission
requirement. To ensure the continuity of the mission, the
design of the overall onboard image processing chain, from
the acquisition by the sensor up to the transmission of the
image data to the ground, must prevent any risk of a bottleneck
that would result in data loss because of memory overflow.
As an example, the CHIME nominal image processing
chain will implement, in the data processing unit (DPU), real-time
cloud detection [168], feeding a hyperspectral Consultative
Committee for Space Data Systems (CCSDS)-123
compressor [169]. This will reduce the amount of data delivered
by the sensor while compressing data with tunable
losses over cloud-free areas [170].
A future unit implementing hypothetical onboard AI algorithms
will have to process data on the fly at the sensor
data rate to interface with the continuous data flow from
the video acquisition unit (VAU). This imposes new constraints
on the architecture of the current onboard data
processing chain (i.e., a new interface with the existing
VAU and updates in the DPU design) as well as strong constraints
on the AIU hardware, such as high-speed input-
output memory links, fast data processing cores, and the
parallelization of DPUs to allow the AIU to process data at a
rate compatible with the sensor throughput while ensuring
the constant monitoring required by the CHIME mission.
INTUITION-1 CASE STUDY
BACKGROUND
The purpose of the Intuition-1 space mission (Figure 3) is
to observe Earth by using a 6U nanosatellite equipped with
a hyperspectral optical instrument and multidimensional
data processing capabilities, which employs deep CNNs.
The mission is designed as a technology demonstrator, allowing
for verification of various in-orbit data processing
applications, use cases, and operations concepts. As of the
third quarter of 2022, the Intuition-1 payload was qualified
for space applications [technology readiness level (TRL) 8],
with a launch scheduled for the second quarter of 2023.
FIGURE 3. The Intuition-1 satellite.
26
MISSION OBJECTIVES
The main objective of the mission is to test a system composed
of in-house-developed components-a high-performance
computing unit, hyperspectral optical instrument,
and data processing software (preprocessing, segmentation,
and classification algorithms)-and evaluate the
applicability of nanosatellites to performing mediumresolution
hyperspectral imaging coupled with onboard
data processing. The important goal of the mission is to assess
real-life advantages and shortcomings of the onboard
data processing chain, taking into account the full mission
context, from the technical perspective (data, power, and
thermal budgets) to spacecraft operations and scheduling.
Intuition-1 will be equipped with a smart compression
component that will prioritize the downlink and processing
pipelines, based on the cloud cover within the area of
interest. The mission is aimed to be multipurpose, with inorbit
update capabilities (through uplinking updated AI algorithms)
and targeting new use cases emerging during the
mission; therefore, the satellite can be considered a " flying
laboratory. " The first use case planned for Intuition-1 is the
onboard estimation of soil parameters from acquired HSIs.
Finally, hyperspectral data captured by Intuition-1 could be
IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE JUNE 2023
KP LABS/AAC CLYDE SPACE
IEEE Geoscience and Remote Sensing Magazine - June 2023

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - June 2023
