HYPSO (HYPerspectral Smallsat for Ocean Observation)
HYPSO is a CubeSat mission of NTNU (Norwegian University of Science and Technology), Trondheim, Norway. HYPSO will be launched into a sun-synchronous polar orbit to observe ocean color along the coast of Norway. Its specific mission is to detect and characterize ocean color features such as algal blooms, phytoplankton, river plumes. etc. The spacecraft will be a 6U CubeSat structure, provided by NanoAvionics LLC. The CubeSat is equipped with a hyperspectral pushbroom imaging payload (hereafter called HSI) which has on-board processing capabilities. Due to thr form of the HSI payload, it requires a 6U satellite bus, but the HSI payload does not occupy the full volume. 1)
Figure 1: The HYPSO-1 mission will observe oceanographic phenomena by using a small satellite with a hyperspectral camera onboard (image credit: NTNU)
Motivations for employing this approach on a nanosatellite are:
• Oceanographic phenomena are of great interest to understand more about the effects of climate change and human impact on the world
• Traditional Earth-Observation satellites are very expensive and take several years to develop and launch
• Dedicated small satellites can be used to provide high spatial resolution within a small field of view to areas of interest with short revisit times
• The information from these images can be downloaded and communicated to unmanned aerial surfac and underwater vehicles which can then investigate the areas of interest further using the data from the small satellite.
There is currently a team of approximately 20 students and 10 PhDs and 2 Post.Docs.researchers working on this mission. The team is multidisciplinary, consisting of students from several departments at NTNU, mainly from the Department of Electronic Systems and the Department of Engineering Cybernetics, in addition to Mechanical Engineering. This is the first satellite for the students and the first chance to work in a truly multidisciplinary team. The project kicked off on 23 August 2018, and will have its PDR (Preliminary Design Review) on 30 October 2019. 2)
Figure 2: The Hyper Spectra Imager (HSI) mission will observe oceanographic phenomena by using a small satellite with a hyper-spectral camera onboard. One use case is monitoring of harmful algal blooms for protection of fish farms (video credit: NTNU)
The mission objectives of the HYPSO-1 are to monitor the spatio-temporal extent of ocean color events in the visual and near-infrared (VIS-NIR) wavelengths between 400 - 800 nm; and to infer phytoplankton functional groups. Key user needs in the ocean color remote sensing are:
1) Images should have spatial resolution better than 30 -100 m per pixel
2) Raw hyperspectral data should have spectral resolution of about 5 nm for VIS-NIR wavelengths
3) The imager’s SNR at Top of Atmosphere (ToA) should be greater than 400 in visual wavelengths for open ocean water , and atmospherically corrected SNR of water-leaving signals should be between 40 - 100
4) Data latency should be less than 1 hr
5) Revisit times to dedicated areas of interest should be 3 -72 hrs.
Since HYPSO-1 is a single small-satellite, but the first in a prospective constellation, we focus on working towards the recommendations 1), 2), 3) and 4).
Image Acquisition Basics
Whereas several types of spectrometers can be integrated on aerial or space platforms, the passive pushbroom imager design is an attractive choice with good SNR. Use of COTS components have also made this type of design more affordable, accessible and flexible.
With the scan direction oriented towards the velocity direction, a pushbroom imager sequentially scans several lines, Nx, each having instantaneous pixels, Ny and Nλ, forming a hyperspectral datacube shown in Figure 3. Ny is the number of spatial pixels perpendicular to the scan direction and Nλ is the number of pixels along the spectral dimension. The horizontal and vertical components of the FoV are εw and εh, respectively. The time elapsed between two consecutive lines, or frames, is expressed by the integration time Δt = 1/F P S = τ + δt where F P S is the frame rate, τ is the camera exposure time and δt is the read-out time.
High spectral resolution is required for discrimination of fine spectral features in the water-leaving signals, and high spatial resolution is desired to reduce the effects of spectral mixing or blur in the image pixels. Mounted on a satellite moving at high orbital speed, the drawback is strictly speaking a low spatial resolution along the scan direction. A work-around is to overlap more frames by tilting the imager backwards as it translates forward, similar to the method described in [3)]. This results in an increased number partially overlapping pixels which can be utilized to enhance SNR or spatial resolution as trade-offs using image restoration techniques such as deconvolution or super-resolution [4)]. For clarity, the Euclidean distance on ground between center pixels of two consecutive frames is taken to be the Sequential Ground Sampling Distance (SGSD) not to be confused with the commonly defined Ground Sampling Distance (GSD) between adjacent pixels in an instantaneous frame.
Concept of Operations
The overall mission utility and performance in HYPSO-1 is mainly engineered based on trade-offs in spatial resolution, spectral resolution, SNR, data size and latency, coverage to ground stations and likely locations for observations. HYPSO-1 will be launched to a 500 km altitude Sun-Synchronous Orbit (SSO) with Local Time of Descending Node (LTDN) at 10:00 AM, which grants early-day access to observe the Norwegian coastline during Spring and Summer seasons while avoiding detrimental sun-glint effects. The HYPSO-1 mission concept of operations (CONOPS), illustrated in Figure 4, enables five main capabilities:
1) After receiving telecommands and updates (e.g. camera settings) that are uploaded from a nearby ground station, HYPSO-1 is scheduled to orient its hyperspectral imager to start scanning a pre-defined area size;
2) HYPSO-1 executes a single-axis slew maneuver so that the imager’s footprint slowly rotates backwards with respect to the scan direction. At a high camera frame rate, the goal is to enable a SGSD better than 100 m/pixel.
3) After imaging, the hyperspectral images are processed onboard immediately to reduce their data size and speeding up the download on ground;
4) For quick downlink after observing coastal regions in Norway, the selected ground station network includes S-band ground stations at NTNU Trondheim, KSAT Svalbard, Norway, and KSAT Puertollano, Spain;
5) In addition, the Mission Control Center at NTNU operates several supporting robotic assets, such as UAVs (Unmanned Aerial Vehicle), ASVs and AUVs, that may collect in-situ data if within range of the observed area.
Figure 4: CONOPS where 1) HYPSO-1 receives uplinked configurations from a nearby ground station; 2) acquires hyperspectral images for a short duration under a slew maneuver; 3) processes the images onboard immediately; 4) downlinks the data to nearby ground stations; and 5) in-situ assets in the vicinity may be deployed for closer investigation at the observed scene (image credit: NTNU)
1) Imaging Modes: The hyperspectral imager has three main imaging configurations:
• High-resolution mode: enables high image resolution with narrow-FoV and high frame rate settings;
• Wide FoV mode: enables a wider swath but at coarser spatial resolution;
• Diagnostics mode: gives raw data with full sensor resolution to be mainly used for in-orbit calibration and characterization.
2) Attitude Determination & Control System: To obtain a spatial resolution better than 100 m requires a precise attitude determination and control system (ADCS). Throughout image acquisition for a satellite that is pointing or maneuvering, the attitude sensor noise and actuator inaccuracies (e.g. reaction wheel jittering) will contribute to a non-uniform distribution of images across the observed scene. The attitude errors are categorized as attitude control and knowledge accuracies, bearing in mind that performance of latter affects the former. For consistent image registration, or simply knowing the location of each pixel to the accuracy of 100 m on ground, e.g. geo-referencing, then good performance is needed for attitude knowledge accuracy, orbit position accuracy, and time synchronization between the captured images and attitude data.
3) On-board Image Processing: The image processing architecture should be modular by design with the goal to ease satellite operations and provide tailored data to end users at a low data latency. To make such data products useful, the high-level goals are to:
- Reduce hyperspectral data size onboard to improve data latency, by lossless compression at minimum;
- Extract the spatial and spectral information in water-leaving signals, by e.g. dimensionality reduction, target detection or classification;
- Register images and utilize the obtained SGSD to achieve better than 100 m/pixel image resolution using image restoration methods, e.g. deconvolution or super-resolution;
- Transform pixel indices to geodetic latitude and longitude by geo-referencing such that these coordinates can be used to guide in-situ agents towards interesting locations.
In the commissioning phase, the hyperspectral data products shall be analyzed in synergy with other available remote sensing data and in-situ measurements. Modeling and simulation tools shall also provide supporting information on atmospheric correction and the radiometric, spectral and spatial properties of a simulated ocean color even.
Launch: The HYPSO-1 satellite was launched on 13 January 2022 (at 15:25 UTC) on the Transporter-3 dedicated rideshare mission of SpaceX on a Falcon 9 Block 5 vehicle from SLC-40 at Cape Canaveral Space Force Station in Florida. On board this launch were 105 commercial and government spacecraft, including CubeSats, microsats, PocketQubes, and orbital transfer vehicles. 5) 6)
Orbit: Sun-synchronous Orbit, nominal altitude of 525 km, inclination of ~98º.
Sensor compliment (HSI)
Note: The HYPSO-1 mission will only feature the HSI payload. This payload consists of the HSI camera supported by an RGB camera for reference. The payload has onboard processing capability on a Xilinx Zync 7030 platform where the HSI-images can be pre-processed before a download. 7)
HSI (Hyperspectral Imager)
HSI is a pushbroom hyperspectral imager with wavelengths of 387–801nm at 3.33 nm bandpass and a swath width of 70 km.
An optical diagram of the instrument with its cross-section parallel to the refraction axis is shown in Figure 5. The labelled components are: (i) front lens with aperture diameter D0 and focal length F0; (ii) entrance slit with width ωslit and height hslit; (iii) collimator lens with aperture diameter D1 and focal length F1; (iv) grating receiving incoming light at angle α = 0º and diffracts the light at angle β measured from the grating normal; (v) detector lens with aperture diameter D2 and focal length F2; and finally (vi) is the image sensor.
Figure 5: Optical diagram of the cross-section of the pushbroom hyperspectral imager. The light is focused into a slit, collimated into a grating which then diffracts the light into an image sensor plane (image credit: NTNU)
HYPSO-1’s hyperspectral imager payload is mainly built with COTS products from Thorlabs and Edmund Optics and a few custom machined parts as shown in Figure 6. The imager is designed to provide a spectral range of at least 400-800 nm and theoretical spectral bandpass of 3.33 nm. The focal length for each lens should be equal to maximize the light throughput, but to avoid detrimental stray light effects the F-numbers are set to F0/# = F1/# = 2.8 and F2/# = 2.
Table 1: Hyperspectral imager specifications
A SONY IMX249 image sensor is mounted in an industrial camera head from The Imaging Development Systems Europe GmbH. It has 1936 x 1216 pixels with reported well depth of about 33022 e- per pixel equivalent to maximum SNR of approximately 181.6. The maximum FPS is limited by the data throughput, number of binning operations, subsampling and Area of Interest (AoI), where latter is the selected number of pixels in a custom image sensor window.
Satellite bus of the HYPSO-1 system
The hyperspectral imager was chosen to be adapted to the Multipurpose 6U Platform (M6P), a commercially available spacecraft bus provided by NanoAvionics, with mass of approximately 6.8 kg when fully integrated. Among the important subsystems in M6P are Flight Computer (FC) for onboard data handling and ADCS functions, SatLab Global Navigation Satellite System (GNSS) for orbit determination and time synchronization, Electrical Power System (EPS), Ultra-High-Frequency (UHF) radio for basic space-ground communications, and Payload Controller (PC) working as storage device and router between the payload and the satellite bus. For internal communications, the spacecraft uses the CubeSat Space Protocol (CSP) over a Controller Area Network (CAN), where each subsystem is a network node with dedicated CSP address. The M6P has 16 body-mounted triple junction Gallium Arsenide solar cells and six Lithium-Ion batteries with total energy capacity of 64.9 Wh.
To fulfill the user needs and mission CONOPS, HYPSO-1 is further equipped with:
• A Nano Star Tracker ST-1 [8)] and Sensonor STIM 210 Inertial Measurement Unit (IMU) [9)] used for precise attitude estimation during imaging. To ensure sufficient settling time after initialization, the sensors are turned on for at least 5 min prior to imaging. When images will not be taken, then six sun sensors, three magnetometers and three gyroscopes are used instead which provide coarser attitude knowledge but consume less power;
• Four reaction wheels used for attitude control that provide up to 3.2 mNm torque each, where three are placed orthogonally along the body axes and the fourth is tilted at an angle of 54.7◦. Two magnetorquers are placed along each body axis for reaction wheel momentum dumping;
• An IDS UI-125x RGB camera with 6 mm F/1.4 Ci series fixed lens providing a footprint of 770 km x 540 km and spatial resolution of approximately 500 m. Its main purpose is to support and validate hyperspectral images in the spatial domain [10)];
• A 2.4 GHz IQ Spacecom S-band Transceiver providing usable data rate of 1 Mbit/s for downlinking payload data;
• An Onboard Processing Unit (OPU) hosting a Zynq 7030 Xilinx PicoZed System-on-a-Chip (SoC) with flight heritage [11)]. It consists of two core ARM processors and a Field Programmable Gate Array (FPGA) dedicated for onboard image processing. The OPU allows for in-orbit updates of both software and FPGA hardware reconfigurations for uploaded algorithms. Larger data sizes can be buffered from the OPU to the PC over CAN before downlinking over S-band radio, or smaller amounts of data can be downloaded directly from the OPU. Buffering data to the PC enables full utilization of the S-band data rate, and removes the need for keeping the OPU turned on for longer than necessary. Power and data-line distribution to the hyperspectral and RGB cameras are granted through a custom break-outboard with PicoZed interfaces. Furthermore, the OPU hosts a SD-card with 8 GB storage capacity.
Figure 7: Computer-Aided Drawing (CAD) model of HYPSO-1 with its top and front panels removed showing the hyperspectral imager in the center, RGB camera to its left and star-tracker to its right (image credit: NTNU)
M6P’s solar arrays generate approximately 11.65 W during a period of 58.9 minutes in sunlight out of a total orbital period of 94.6 minutes. Determining if energy is sufficient during burdensome operations, the power budget should assume a scenario where image acquisition, processing and downlink all happen in the same pass during sunlight. This scenario is shown in Figure 8 for HYPSO-1 passing over a target area in Lofoten, Norway, and the selected ground stations at NTNU Trondheim, KSAT Svalbard and KSAT Spain.
Figure 8: HYPSO-1 in SSO at 10:25:00 on 28 May 2022. Selected ground stations are marked in white circles. Previous, current and succeeding ground tracks are indicated by dashed lines (image credit: NTNU)
Overview of Onboard Image Processing Architecture
The FPGA-based image processing algorithms on the OPU are key to enable faster download and distribution of data while at the same time relieving HYPSO-1’s power budget. The idea behind the image processing architecture is to allow for modular arrangement of algorithms, or pipelines, as illustrated in Figure 9. The minimal, dimensionality reduction, target detection, and classification on-board image processing pipelines (respectively named MOBIP, DROBIP, TOBIP and COBIP) are designed to generate tailored data products depending on the particular need of the user or operator. All pipelines include image acquisition, time-stamping and binning prior to image processing. It is critical that also satellite and payload telemetry and any other relevant metadata are downlinked together with the processed images, including ADCS and orbit position data collected during image acquisition.
Figure 9: Block diagram of the proposed imaging processing pipelines. In order, the hyperspectral images are captured, binned, processed at chosen level, stored on SD-card and downlinked together with collected telemetry and metadata. Depending on the downlinked data product, additional ground-based processing and fine-tuning can be applied before distribution to end user. In-situ validation and data fusion with other remote sensing data are critical aspects the HYPSO-1 data validation. Black arrows indicate the minimal onboard processing pipeline while gray arrows are the alternative routes for tailored data products (image credit: NTNU)
Table 2 shows the size reduction and processing speeds for the suggested algorithms to be employed in the architecture. ”Bands/Components” are referred to as spectral bands for raw data and MOBIP, extracted components for DROBIP, a probability map of detected target spectral signatures for TOBIP, and a layer containing classes of spectra for COBIP. A raw hyperspectral datacube of 956 x 684 spatial pixels and 1080 spectral pixels binned by a factor of Bλ = 9 times is considered as the starting point for further processing. The data size reduction and processing speed estimates are based on performance reported on state-of-the-art image processing algorithms that have been used on hyperspectral data of similar sizes. Details related to occupation, execution time, operating frequency and latency of the following FPGA-based algorithms can be found in the respective literature on their implementation.
Table 2: Hyperspectral data products for Nx x Ny = 956 x 684 spatial pixels (Ref. 2)
Sensor complement for the HYPSO-2 mission (HSI, SDR)
Note: The HYPSO-2 mission is planned for launch in 2023. It will carry both instruments, HSI and SDR.
Figure 10: Photo of the SDR and HSI payloads for the HYPSO-2 mission (image credit: NTNU)
SDR (Software Defined Radio)
In the ground segment there will be a ground station network and sensor node prototypes for the future Arctic communication system (Figure 11). The S-band ground station will be used as main TT&C (Telemetry, Tracking and Command) for the HSI, and to downlink interference measurement data. The UHF ground station is a backup for TT&C and it is also used to perform downlink channel measurements and transmit reference signals for calibration. Sensor nodes prototypes for future Arctic communications will be used to do channel measurements for the use case and to demonstrate a communication link.
The space segment is formed by the satellite. The SDR payload will measure both radio interference and communication channel. It will also demonstrate a communication link with sensor nodes prototypes. The S-band communication components will be used for the same as the S-band ground station. The UHF radio and turnstile antenna will be TT&C backup. The SDR payload will use the UHF monopole antenna for the measurements to avoid interfering with main communications or data link of the spacecraft. The SDR payload must communicate with the payload controller of the satellite bus to downlink data through S-band and get navigational data.
The main constraints for the design of the payload are cost and development time. Schedule constraints are very important in the trade-offs for the secondary mission to be compatible with HYPSO project.
A total of 21 SDR platforms have been analyzed and have been part of a high level assessment. 12) An extra alternative was found after that study, TOTEM SDR from Alén Space. Power consumption is quite low compared to the alternatives, it includes the Radio-Frequency (RF) front-end and its noise figure is 2 dB. The transceiver chip has only one transmitter and one receiver chain. Nevertheless, as cost is reasonable, and it provides high level of flexibility it was decided that this platform will be the SDR payload of the mission. Since SDR-UR-006 states that the frequency band should be between 400-440 MHz, but the front-end filters have a bandwidth of 10 MHz, a bypass was included. Signals in this branch (additional RF I/O in the picture) will not pass through the filters and amplifies of the front-end. This was the only solution found to avoid connecting another front-end board.
Table 3: TOTEM characteristics
In Figure 13 the architecture of the TOTEM platform and how it can be connected to the antenna is shown. This platform is formed by two boards: RF front-end (analog part) and SDR motherboard (analog stage, analog/digital conversion and digital processing). The SDR motherboard consists of an RF transceiver (AD9364) and a System on Chip (SoC) based on Xilinx boards, which has a Zynq 7020.
Mass/volume: The volume of the spacecraft is 6U, leaving room for the SDR payload in conjunction with the HSI payload. Because the SDR payload radio does not require much mass nor volume, the constraints imposed by the HYPSO mission do not influence the radio module itself. Except for the choice of antenna and antenna placement, described in the next section. The SDR radio has masses that influence the spacecraft's moment of inertia and center of gravity, but the internal configuration and the arrangement of subsystems within the spacecraft do not influence the mission significantly.
In addition, a mechanical interface for TOTEM is required. The SDR has a PC104 form factor, but due to the placement of the HSI and other components in the bus, the SDR has no available space to be mounted on stacking rings used for PC104. Therefore, an alternative mounting assembly had to be designed. The custom hardware interface (Figure 3) consists of: mounting plate, base plate as a platform for mounting, cylinder spacers to extend the support from the base plate to the SDR and provide a stable base and a support plate to provide support for the rods and reduce the moment that the SDR may impact on them.
The SDR mission designers must work closely to ensure transparent and up-to-date communication with the HYPSO spacecraft designers not to compromise the main mission of the spacecraft. Thus, a mass budget for the secondary payload is required. The payload mass budget of the SDR payload is shown in Table 7. The UHF monopole antenna is not included in the payload budget as it is included in Nanoavionics satellite bus.
Table 4: Payload mass budget
Antenna: The HYPSO mission is equipped with two imaging payloads that need a specific FOV (Field of View) to operate. These parameters give the main constraint on the antenna design for the SDR: SDR antenna placement shall not interfere with any of the imaging payloads. The FOV of the HSI is assumed to be ±4.22° and the RGB camera has a FOV of ±35°. The HSI will be placed in the middle of the 2U side of the satellite (3U axis aligned with Earth radius) and the RGB in the middle of one the 1U in the same side.
The satellite bus has three antennas: one S-band patch antenna, one UHF turnstile and one UHF monopole antenna. For channel measurements a turnstile antenna with an omnidirectional pattern would be desired to easily distinguish the effect of the antenna pattern from the channel or interference effects. However, the turnstile antenna in the bus is used for communication during Launch and Early Orbit phase (LEOP) and as a backup for TT&C. Thus, the SDR can only utilize the UHF monopole which may only be deployed if it does not interfere with the FOV of the imagers.
Figure 14 shows a placement of the antenna to get compromise between an omnidirectional antenna pattern and camera FOVs. Assuming a 15 cm monopole, the antenna must be placed so that Δx1> 1.1 cm and Δx2>10.5 cm, shown in Figure 14. The monopole will be placed 11 cm from the center of the RGB camera.
Challenges as a secondary payload: If a secondary payload is added to the satellite after the satellite bus is selected, this payload must be adapted to the bus. The most important requirement for a secondary payload in this case is to limit the impact on the primary mission. This must be ensured during integration, thermal analysis, system budgets and testing.
KSAT has signed a contract for ground station support of the HYPSO CubeSat missions from the Norwegian University of Science and Technology (NTNU), aiming to detect toxic algae blooms — this is the first time KSAT will provide ground station services to a Norwegian university mission. 13)
At NTNU Small Satellite Lab, a multi-disciplinary team of master students, PhD-students and professors are currently working on a small satellite with a miniaturized hyperspectral camera for detection of toxic algae blooms along the Norwegian coast. KSAT will as part of this contract, provide commercial ground station services from the Svalbard Ground Station for this mission, called HYPSO-1.
The smallsats in the HYPSO-project will be equipped with miniaturized hyperspectral cameras, that are able to “see” more than traditional optical sensors, covering parts of the infrared spectrum. In combination with drones and autonomous vehicles both on surface and subsea, the goal is to be able to detect and alert the fish-farms about toxic algae blooms in the area. In 2019 a sudden upwelling of toxic algae killed close to 8 million salmon in Norwegian fish farms, wiping out more than half of the annual sales growth in just over a week. The hope is that with the contribution of this mission, one can avoid this in the future.
As a significant provider of maritime monitoring services, KSAT had an active role during the algae bloom last year and together with partners in Tromsø they are currently exploring how to discriminate between different types of algae by combining different sensors and applying advanced algorithms.
“We are very excited to get access to KSAT's ground stations both at Svalbard and at other locations,” said Associate Professor Egil Eide at the Department of Electronic Systems. “HYPSO-1 will be part of a multi-agent surveillance system, operating both drones and surface vessels in near-real-time. It is very important to get data from as many satellite passes. This is an important strategic cooperation between NTNU and space industry, that will benefit students and researchers alike.”
Kristian Jenssen is the Director of KSATLITE, a division at KSAT that is dedicated to the development and delivery of scalable, global ground station services for smallsats. The team are currently handling the major portion of the commercial smallsats on-orbit today, including some of the large constellations. Jenssen emphasized that the students through these projects acquire unique hands-on experience, which is very relevant and thus valuable for KSAT as a possible future employer. “It is important for us that students that want to delve into the discipline of spacecraft engineering and space related sciences, can get the chance to do so at Norway’s largest technical university,” stated Jenssen, adding, “It’s exciting with these new and small hyperspectral sensors. We support the project and believe it can provide a valuable contribution to future systems for algae warning and coastal monitoring to increase the understanding and support commercial aquaculture.”
Figure 15: KSAT's Svalbard ground station (image credit: KSAT)
1) Hyper Spectral Imager for Oceanographic Applications,” NTNU, URL: https://www.ntnu.edu/ie/smallsat/mission-hyper-spectral-camera
2) Mariusz E. Grøtte, Roger Birkeland, Evelyn Honor é-Livermore, Sivert Bakken, Joseph L. Garrett, Elizabeth F. Prentice, Fred Sigernes, Milica Orlandi ć, J. Tommy Gravdahl, Tor A. Johansen, ”Ocean Color Hyperspectral Remote Sensing with High Resolution and Low Latency – the HYPSO-1 CubeSat Mission,” IEEE Transactions on Geoscience and Remote Sensing, Vol. X, No. X, February 2021, URL: https://folk.ntnu.no/torarnj/HYPSO_Concept_Paper__Post_Submitted_Rev_3_.pdf
3) D. Jervis, J. McKeever, B. O. A. Durak, J. J. Sloan, D. Gains, D. J. Varon, A. Ramier, M. Strupler, and E. Tarrant, “The GHGSat-D imaging spectrometer,” Atmospheric Measurement Techniques,” Volume 14, pp:2127-2149, 2021, URL: https://amt.copernicus.org/articles/14/2127/2021/amt-14-2127-2021.pdf
4) Simon Henrot, Charles Soussen, David Brie, ”Fast positive deconvolution of hyperspectral images,” IEEE Transactions on Image Processing”, Vol. 22, No. 2, pp. 828–833, Feb. 22, 2013, https://ieeexplore.ieee.org/document/6291782
6) Jeff Foust, ”SpaceX launches third dedicated smallsat rideshare mission,” SpaceNews, 13 January 2022, URL: https://spacenews.com/spacex-launches-third-dedicated-smallsat-rideshare-mission/
7) Information provided by Roger Birkeland of NTNU
8) Yu-ming Li, Chun-jiang Li, Ran Zheng, Xiao Li, Jun Yang, ”The research on image processing technology of the star tracker,” Proceedings of SPIE, Volume 9301,' International Symposium on Optoelectronic Technology and Application 2014: Image Processing and Pattern Recognition; 930103 (2014),' https://doi.org/10.1117/12.2072128 , Event: International Symposium on Optoelectronic Technology and Application 2014, 2014, Beijing, China
9) T. S. Rose, D. W. Rowen, S. LaLumondiere, N. I. Werner, R. Linares, A. Faler, J. Wicker, C. M. Coffman, G. A. Maul, D. H. Chien, A. Utter, R. P. Welle, S. W. Janson, ”Optical communications downlink from a 1.5U Cubesat: OCSD program,” Proceedings of SPIE, Volume 11180, 'International Conference on Space Optics — ICSO 2018; 111800J (2019),' https://doi.org/10.1117/12.2535938, Event: International Conference on Space Optics - ICSO 2018, 2018, Chania, Greece
10) Ayman Habib, Youkyung Han, Weifeng Xiong, Fangning He, Zhou Zhang, and Melba Crawford, ”Automated Ortho-Rectification of UAV-Based Hyperspectral Data over an Agricultural Field Using Frame RGB Imagery,” Remote Sensing, Published: 24 September 2016, Vol. 8 (10). 796, https://doi.org/10.3390/rs8100796
11) Alan D. George, Christopher M. Wilson, et al., ”Onboard Processing With Hybrid and Reconfigurable Computing on Small Satellites” Proceedings of the IEEE, Volume: 106, Issue: 3, March 2018, pp: 458-470, Publication: 27 February 2018, DOI: https://doi.org/10.1109/JPROC.2018.2802438
12) G. Quintana-Diaz and R. Birkeland, ”Software-defined radios in satellite communications,” 4S (Small Satellites, Systems & Services) Symposium, Sorrento, Italy, May 28-June 1, 2018,
13) ”KSAT To Support HYPSO-1 Smallsat With Ground Station Support,” Satnews, 11 August 2020, URL: https://news.satnews.com/2020/08/11/ksat-to-support-hypso-1-smallsat-with-ground-station-support/
The information compiled and edited in this article was provided by Herbert J. Kramer from his documentation of: ”Observation of the Earth and Its Environment: Survey of Missions and Sensors” (Springer Verlag) as well as many other sources after the publication of the 4th edition in 2002. - Comments and corrections to this article are always welcome for further updates (email@example.com).