The Earth observation sector continues to evolve at a fast pace. With rapidly falling costs, reductions in risk, and new launch technologies, there has been a surge of new satellites and capabilities entering the commercial market. Multiple low-cost satellite constellations, both optical and synthetic aperture radar, are being launched to support data collection at high frequency; and data analytics companies aiming to take this data and develop information solutions.
Most of these new constellations are focusing on the visible channels — red, green, blue — with perhaps a band or two in the near-infrared (NIR) to support vegetation monitoring and land-use classifications. Few of them are looking to Hyperspectral sensors that collect image data in hundreds of contiguous narrow spectral bands and can provide detailed information about target spatial and spectral patterns. Hyperspectral payloads are not new, they have been mounted on aircraft for decades, with the earliest commercial unit being the Geophysical Environmental Research Imaging Spectrometer II, developed by the US and launched in 1987 and also on US Predator in past.
Spectroscopy technology or spectrometry deals with the measurement of a specific spectrum for identification of matters. It is a key analytical method used to investigate material composition and related processes through the study of the interaction of light with matter. The energy is absorbed by the matter, creating an excited state. The interaction creates some form of electromagnetic waves. By using a spectrometer, one can determine the level of excitement in the matter’s atoms to determine what kind of material it is. Determining composition remotely, without physical contact, is one of the most valuable capabilities of spectroscopy.
An imaging system converts the visual characteristics of an object, such as a physical scene or the interior structure of an object, into digital signals and creates digitally encoded representations that are processed by a processor or computer and made output as a digital image. Imaging systems typically consist of a camera, imaging lens, along with an illumination source. Depending on the system setup, an imaging system can allow observed objects to be magnified or enhanced to ease the viewing or inspection of small or unclear objects. Computers are becoming more and more powerful with increasing capacities for running programs of any kind especially digital imaging software.
The combination of spectroscopy technology and the modern imaging system is referred to as imaging spectrometry, now also called hyperspectral imaging. It could measure a spectrum for every element (or pixel) in an image. This provides a revolutionary way of observing the earth and other planets by collecting information of each pixel in the field of view across the electromagnetic spectrum.
A hyperspectral imager operating in the solar reflected spectrum senses objects in the field of view in detail spectrally and spatially. Molecules and particles of the land, water, and atmosphere environments interact with solar energy in the 400–2500 nm spectral region through absorption, reflection, and scattering processes. These spectral measurements are used to determine constituent composition through the physics and chemistry of spectroscopy for scientific research and applications over the regional scale of the image.
The main advantage to hyperspectral imaging is that, because an entire spectrum is acquired for each pixel of the acquired imagery, an operator needs no prior knowledge of the sample, and postprocessing allows all available information from the dataset to be exploited. Hyperspectral imaging can also take advantage of the spatial relationships among the different spectra in a neighborhood, allowing more elaborate spectral-spatial models for a more accurate segmentation and classification of the image
HSI systems normally operate from within the visible spectrum of light – between 390-700 nanometres (nm) – up to a point in the long-wavelength (infrared) at 15,000 nm. Although commercial sensors are available for the entire range, they tend to be comparatively bulky and include moving parts unsuitable for use in the microgravity of space.
Almost all the spaceborne hyperspectral imagers use 2-D area detector arrays and operate in pushbroom operating mode. As shown in the figure, it images an entire line of ground sampling cells in the cross-track direction, whereas an aircraft or spacecraft provides the forward scan in along-track direction. The 1-D image of the cross-track line, formed on the spectrometer slit, is then dispersed onto the 2-D detector array, which provides spectral information along one axis and spatial information along the other. This architecture effectively integrates as many individual spectrometers as there are ground sampling cells in the cross-track line into a single instrument.
The advantages of a dispersive element based hyperspectral imager that operates in the pushbroom mode are as follows.
- No moving parts.
- Congruence spatial images.
- Longer integration time for each ground sampling cell, because each of them is sensed simultaneously by a row elements of the 2-D detector array (e.g., rows A, B, C,…, G in Fig. ) instead of one after another, which omits the time sharing scanning of all the ground sampling cells in a cross-track line. Longer integration time means more photos are collected and results in higher signal-to-noise ratio (SNR).
The disadvantages of a pushbroom hyperspectral imager are as follows.
- Complex optical design and complex focal plane.
- Swath width is constrained by the available number of pixels of the 2-D detector array in the spatial direction.
- Complex calibration.
- There are both spectral distortion (also referred to as smile) and spatial distortion (also referred to as keystone). The hyperspectral data collected by a pushbroom hyperspectral imager need to be sufficiently corrected for smile and keystone distortion before being distributing to users for downstream applications
- A primary advantage of hyperspectral remote sensing image data is capability to discriminate, classify, identify as well as quantify materials present in the image. Absorption and emission bands of given substances often occur within very narrow bandwidths. This allow high-resolution, hyperspectral sensors to distinguish the properties of the substances to a finer degree than an ordinary broadband sensor. The intensity of this energy can be measured at various wavelengths. Many objects and substances have spectral characteristics that are unique and a unique spectral “signature” allows them to be identified through various spectral analyses. The target spectral signatures are also different from the background signature.
- Passive spectral sensors have the potential for detecting and characterizing camouflaged and concealed targets in clutter. By analyzing the multiple spectrum bands of HSI it is possible to detect concealed targets by keying on their reflectance differences at non-visual wavelengths.
- It also allows one to detect targets of interest with sizes smaller than the pixel resolution (sub pixel target detection), and abundance estimation, which allows one to detect concentrations of different signature spectra present in pixels.
Analysts can compare collected sensor data with libraries of known material signatures, enabling classification. Software featuring these techniques and capabilities is already available in the commercial sector from companies such as Harris Geospatial, Hexagon Geospatial, and BAE Systems.
Limitations of HSI
- HSI’s shortcoming is that its spatial resolutions are usually coarser than those of panchromatic imagery due to the trade-off with fine spectral resolution. A panchromatic HRI image, typically with much better spatial resolution, provides additional spatial enhancement on target. Image visualization can be enhanced by spatially sharpening a hyper spectral image with a panchromatic high-resolution image.
- Since hyperspectral technologies are passive, receiving reflected light from surfaces, they are degraded by the frequent cloud cover and fog of tropical environments. In addition, water vapor, oxygen and other gases differentially attenuate and change the reflected energy spectrum affecting the ability to accurately characterize targets.
The power requirements to support collection across bandwidth ranges at a sufficient ground resolution would lead to a reduced image swath, thus hampering revisit times from a single satellite. Reducing revisit times requires a constellation, with obvious cost implications. Stable platforms are also needed, meaning it is challenging to base this on a lower-cost small satellite design (although this is being explored by Satellogic). Despite the potential limitations on ground resolution, much detail can be garnered from imagery.
Creating an operational satellite system with sufficient ground resolution (less than 10 meters per pixel), good signal-to-noise ratio and decent revisit times, however, has proven to be a challenge. If a ground resolution of sharper than 10 meters and weekly or better revisits can be achieved, then the commercial opportunity for hyperspectral starts to become more interesting. Particularly in the defense realm, there is sensitivity in using hyperspectral data to detect true versus camouflaged objects.
There are three limitations that will have to be overcome before commercial space-based HSI becomes widely available. These are: processing power, communications bandwidth, and power supply. The amount of information collected by a hyperspectral sensor is significantly greater than with conventional electro-optical imagery. As a general rule, an hour of hyperspectral footage would generate about 1 TB of data. Similarly, hyperspectral sensors also require significant processing at the point of collection to make them viable. Given the amount of data produced, communicating this from the satellite to the ground segment using radio frequencies will continue to be a challenge.
Currently, the only solution is to store the data and bring it back to earth for processing, which limits the ability to responsively target the sensor during its mission. As such, at least in the commercial sphere, hyperspectral is not suitable for time-sensitive operations. However, it may be used in agriculture and mining, where longer timeframes are acceptable.
The last limitation is power supply. Infrared imaging systems, including hyperspectral imagers, pose a challenge in this regard, as – beyond a certain wavelength – the heat produced by the electronics will begin to distort the information the sensor collects. In space, where the only way to remove heat is to radiate it, this necessitates the implementation of expensive active-cooling systems and heat tanks. Although advances in solar technology will increase the amount of energy that can be gained from this source, it will remain a limitation, particularly on smaller satellites with smaller solar panels.
There are many commercial and military applications of hyperspectral satellites. They are generally being used to support scientific research, including for example into waterborne pollution levels. Agriculture is also a key area for hyperspectral. Today, satellite-based agriculture applications are based on multispectral solutions with bands (nominally three to five channels) spanning visible red into the NIR (the “red-edge”). By being able to scan the same spectral range in tens or hundreds of bands of more detail, crop health and yield can be assessed. The benefit of having SWIR channels would also support further application development into biophysical and chemical crop properties. There is further applicability to serve the forestry, oil and gas, and environment monitoring sectors.
The new generation of sensors, such as those produced by manufacturers including IMEC, are small, lightweight solid-state sensors weighing less than a few hundred grams, which makes them more practical for use in microsatellites. These new designs facilitate simplified production, which enables mass production and reduced costs. The trade-off is that these newer sensors have a smaller spectral range and are often less sensitive, which at present would limit their usefulness when based on satellites.
For example, in 2016, researchers from Microsoft published a paper demonstrating a low-cost hyperspectral camera that cost less than USD100 to assemble, although a camera of this sophistication would have few space-based applications. There are only a limited number of non-military or government HSI sensors in orbit today.
Military Applications of hyperspectral Imaging
Hyperspectral imaging have ability to observe objects which conceal their emissions in one part of the spectrum like stealth aircraft and thermally suppressed engines or are hidden (such as underground bunkers). They can be a valuable tool for finding submarines and underwater mines in shallow waters.
On land, they can determine the actual composition of objects to distinguish decoys (hyperspectral imaging can capture the differences in EM signature of a wooden decoy versus an actual missile launcher). Enrichment facilities have a visual signature much like other industrial buildings, which makes them difficult to positively identify through visual imagery. These medium-to-large boxy facilities are also easy to bury, hide, and disguise.
Camouflaging is the process of merging the target with the background with the aim to reduce/delay its detection. It can be done using different materials/methods such as camouflaging nets, paints. Hyperspectral satellites are capable to locate and track military targets that are usually camouflaged or hidden underground, such as missile launch sites and testing facilities for nuclear weapons.
HSI is probably already used in the identification of buried roadside bombs or improvised explosive devices (IEDs) and the detection of certain chemical attacks. In 2016, Henrik Petersson and David Gustafsson from the Swedish Defence Research Agency published a paper setting out a series of methods that can be used to detect the presence of IEDs by inspecting disturbed earth using land-based hyperspectral imagery.
In the air, hyperspectral sensors can passively detect even thermally shielded stealth aircraft. For counter-WMD missions, hyperspectral imaging can be used to detect nuclear and chemical weapons production, as well as locating the underground tunnels and bunkers that would house those strategic assets. HSI sensors offer superior spectral resolution that allows stealth aircraft identification even in presence of coatings. However they do not provide a complete solution in themselves as they are degraded by the frequent cloud cover and fog of tropical environments. However they can combine with other sensors to produce better counter stealth solution.
Similarly, Canadian company Telops sells land- and air-based hyperspectral infrared cameras that can detect the presence of methane in the atmosphere. A similar approach could potentially be used to detect the presence of airborne chemical agents.
Hyperspectral imager Hyperion onboard EO-1 satellite is well-known and often regarded as the first spaceborne hyperspectral imager in the remote sensing community. Hyperion is a pushbroom hyperspectral imager using 2-D area detector arrays. It had a relative narrow swath of 7.65 km. The ground footprint size is 30 m × 30 m. The 30 m size in the along-track direction was obtained by basing the frame rate on the velocity of the spacecraft for a 705-km orbit. The entire 7.65 km wide swath is obtained in a single frame. Each image is a data cube for 7.65 km wide in cross-track direction by 185 km long in along-track direction with 242 spectral bands
During a short period from 2016 to 2019, nine spaceborne hyperspectral imagers have been launched into space. There is a leap for the number of spaceborne hyperspectral imagers launched in 2018. Several new spaceborne hyperspectral imagers have been under development for years or have been planned and will come up. In terms of the platforms and orbits of these spaceborne hyperspectral imagers, majority of them (19 of them) are aboard satellites on low earth orbits (LEOs), including three of them deployed on the ISS. Six hyperspectral imagers are out of earth orbits, one (CRISM) on a Mars orbit, one (M3) on a lunar orbit, and one (VNIS) on a lunar rover for in situ observation. The Visible and Infrared Thermal Imaging Spectrometer (VIRTIS) and its two slight variants were deployed onboard the space probes of three planetary missions on orbits of a comet, Venus and two protoplanets.
Regarding the operation mode, among the 25 spaceborne hyperspectral imagers, all of them use 2-D area detector arrays and operate in pushbroom mode. Due to the large data volume generated by hyperspectral satellites, onboard data compression is sometimes adopted to reduce the data volume to ease the data transmission to ground. At least, five spaceborne hyperspectral imagers have used or will use an onboard data compression unit. These are the four hyperspectral imagers on orbits: VIRTIS, CRISM, M3, and HISUI, as well as the one to be launched EnMAP
Many countries have already launched hyperspectral satellites such as European Space Agency’s Sentinel-3 and Flex; the Indian Space Research Organisation’s IMS-1; the Belgian PROBA-1; and PRISMA for the Italian Space Agency. Most missions have a ground resolution of greater than 20 meters, which is less applicable for commercial use; their focus is likely more geared to R&D and scientific usage.
There are, however, several commercial solutions gaining some traction. Montreal-based NorthStar Earth & Space plans a 40-satellite constellation to offer daily revisit with an expected ground resolution of better than 10 meters. It recently completed a $52 million financing round (in addition to the $31 million already achieved) which includes contributions from the federal government of Canada and the provincial Quebec government. Further partners include Telesystem, Telespazio and Thales Alenia Space.
The EnMAP is a German hyperspectral satellite mission scheduled to be launched in 2021. It aims at monitoring and characterizing the earth’s environment on a global scale by providing high‐quality hyperspectral data. EnMAP is a dispersive element (using prisms) based hyperspectral imager operating in pushbroom mode. It has 242 spectral bands covering a wavelength range from 420 to 2450 nm with an SSI of 6.5 nm for VNIR bands and 10 nm for SWIR bands. Its ground swath width is 30 km with a GSD of 30 m × 30 m. It is designed to achieve better SNR than the available spaceborne hyperspectral imagers. The SNR will be greater than 500:1 for a 10 nm equivalent bandwidth of the spectral band at 495 nm. In the SWIR region, an SNR of more than 150:1 will be reached
HyperSat LLC also announced that it has secured an initial $85 million investment from an equity consortium led by Incentrum Group to fund the development of a hyperspectral constellation. The company targets better than 10-meter ground resolution from an initial constellation of six satellites. The first two satellites are expected to be in orbit in 2020. The Satellogic solution targets 30-meter ground resolution hyperspectral data; it announced a $27 million Series B financing round in 2017 with investment led by Tencent.
China Commercial Remote-sensing Satellite System (CCRSS), can collect data on 328 electromagnetic bands, offering very high resolution of up to 15 meters, according to the researchers from the Institute of Remote Sensing and Digital Earth in Beijing. This means each pixel in the image measures 15 metres squared.
In sep 2019, five satellites were launched by a Long March-11 carrier rocket. The satellites belong to a commercial remote-sensing satellite constellation project “Zhuhai-1,” which will comprise 34 micro-nano satellites, including video, hyperspectral, and high-resolution optical satellites, as well as radar and infrared satellites.
The newly launched satellites comprise four hyperspectral satellites with 256 wave-bands and a coverage width of 150 km, and a video satellite with a resolution of 90 centimeters. The Zhuhai-1 hyperspectral satellites have the highest spatial resolution and the largest coverage width of their type in China. The data will be used for precise quantitative analysis of vegetation, water and crops, and will provide services for building smart cities, said Orbita, the largest private operator of hyperspectral satellites in orbit.
TacSat-3 was the first US military satellite with an HSI capability, according to information in open sources. The satellite, which deorbited in April 2012, carried a system called Advanced Responsive Tactically Effective Military Imaging Spectrometer (ARTEMIS) and its associated sensor processor. The Artemis sensors first tested on the TacSat-3 satellite can collect data on 300 electromagnetic bands, thus allowing its user, the US Strategic Command, to operate it for tactical purposes ranging from the detection of roadside bombs to the identification of nuclear weapon facilities.
Italian Space Agency ASI (Agenzia Spaziale Italiana), launched PRISMA satellite that is observing the earth with a hyperspectral optical sensor and collecting data for monitoring and predicting environmental changes on our planet. PRISMA (PRecursore IperSpettrale della Missione Applicativa) is equipped with an innovative electro-optical instrument that combines an hyperspectral sensor with a medium-resolution panchromatic* camera. This combination offers the advantages of conventional earth observation by recognizing the geometric characteristics of a landscape but is additionally able to determine the chemical and physical properties of objects in the landscape through the use of hyperspectral sensors. Roberto Aceti, Managing Director of OHB Italia, said “we are proud to deliver the first European hyperspectral satellite to ASI. This satellite will open new frontiers on services and applications.”
India launched HySIS, a dual use satellite, in November 2013, which is used by the navy. HySIS carries two payloads, the first in the Visible Near Infrared (VNIR) spectral range of 0.4 to 0.95 micrometers with 60 contiguous spectral bands and the second in the Shortwave Infrared Range (SWIR) spectral range of 0.85 to 2.4 micrometres with a 10 nanometre bandwidth and 256 contiguous spectral bands. The satellite will have a spatial resolution of 30 meters and a swath of 30 km from its 630 km sun-synchronous orbit.
The OCI is a hyperspectral imaging radiometer onboard NASA’s Plankton, Aerosol, Cloud, Ocean Ecosystem (PACE) satellite, which is scheduled to be launched in 2022. OCI will be the most advanced ocean color hyperspectral imager in NASA’s history.
The MAJIS has been selected as one of the scientific payloads by ESA for its Jupiter Icy Moons Explorer (JUICE) mission intended to explore Jupiter and three of its icy moons: Europa, Callisto, and Ganymede. It is scheduled to be launched in June 2022. The spacecraft of the JUICE mission is targeted to fly by Callisto, Ganymede and Europa, then a one-year orbital phase around Ganymede
References and Resources also include: