BREMEN, Germany — Companies engaged in hyperspectral satellite imagery are beginning to overcome challenges to unlocking its potential, according to industry experts.
“In recent years technological developments have matured the technology, so now it has become a competitive technology and I think that is what is new,” Kirsten Drost, business development manager at S[&]T, said during a panel at Space Tech Expo Europe in Bremen, Nov. 15.
Optical imagery, which deals with light, is visible to the human eye. Covering limited bands of light, it can reveal shapes on Earth with high spatial resolution. Hyperspectral imagery satellites meanwhile scan tens or hundreds of bands of light allowing for a more nuanced understanding of phenomena on Earth. This includes determining material composition and subtle changes in objects, like the health of crops or quality of water, which might not be visible otherwise.
But there are many challenges, including managing bandwidth tradeoff versus spatial resolution, the sheer volume of data, its complexity, and the laws of physics, and issues of commercialization.
“The technical one is the amount of data,” said Roman Bohovic, chief technology officer at World from Space s.r.o. “Now we are mostly processing data from tens or several tens of bands, but now we need to multiply it by one or two orders of magnitude.”
This high spectral resolution produces vast amounts of data. And getting this data down to Earth, its quality, processing and scaling processing all present challenges.
The industry is looking towards advanced solutions like cloud computing, onboard data processing, and AI to manage this data deluge.
“Onboard data processing is one of the key technologies that might help us to cope with [the vast amounts of data]” Marco Esposito, managing director at cosine Remote Sensing BV, said. This technology is still developing.
AI and machine learning is another way to tackle the issue, helping to decipher the complex patterns in hyperspectral data. Conversely, while there is too much data from the perspective of engineers trying to get data to the ground, there is a dearth of training data for AI.
“We just need more data because this is very experimental at the moment because there are no training data,” says Bohovic. “There are no operational satellites for that. But once we have the use case, that’s not such a big deal to train for this specific use case.”
The practical applications of hyperspectral imagery are vast and varied. These include critical areas like climate change monitoring, agricultural optimization, and environmental protection.
In response to moderator Torsten Kriening, Esposito stated that phenomena like algae and other kinds of pollution can occur very quickly. This can swiftly make water undrinkable, and is a good case for the use of hyperspectral imagery. “If you do not spot that in time this can cause really big damages to the health of people. So people are dying. To recover from some big disasters, millions are spent on recovery when it’s late.”
There is also defense and security sector interest. “A lot of the use cases for the security domain are about detecting camouflage, for instance,” says Drost. The range of bands allows for better distinguishing between decoys or camouflage materials versus actual materials. “But you really need to know spectrum signatures for that. One thing that’s very challenging at the moment is spectral fingerprinting and how to distinguish one material from the other one. Especially in the defense and security domain.”
Hyperspectral is complementary to high-resolution optical imagery, Esposito notes, explaining that the current state of the art for hyperspectral is 30 meter resolution. Companies are looking at one-meter-resolution, but this, and even five or 10 meter resolution imagery for smaller satellites, comes up against physical boundaries. This, he says, means that hyperspectral is best complementing high-resolution optical resolution imagery.
Hyperspectral imagers use a prism or diffraction grating to split the collected light into a range of wavelength bands. The sensors use an array of detectors, each tuned to a different band of light. Each detector can only cover a small part of a scene, constraining spatial resolution.
While the technology is developing, the commercial use of hyperspectral data is still in its nascent stages. Most hyperspectral data providers currently are governmental, and commercial accessibility is limited. Standardized data formats, for example, are also needed to foster a broader commercial adoption.
“Now you see multiple commercial providers who are getting attracted by this new technology and seeing also the potential. But it has taken many years to mature the technology to make it competitive,” Drost says.
As the technology matures, hyperspectral imagery could play an increasing role in remote sensing. This includes areas such as global environmental monitoring, resource management, and sustainable development, as well as in defense.
“The market is dynamic, I would say. I wouldn’t call it saturated,” Drost says. “But there are a lot of missions to be launched. We will need to see how that evolves.”
