Metaspectral delving deeper from above

Open this photo in gallery:

If everything works out according to plan, the computer-vision technology that Francis Doumet and Migel Tissera’s company began working toward in a co-working space in 2018 will be on board the International Space Station next year.

There, the imaging company they founded called Metaspectral will work with American company HySpeed Computing to capture images using a far broader range of light than the human eye can see – wavelengths deep into the infrared spectrum that can help clients and scientists learn far more than a regular picture could show them.

Using information captured of the Earth’s surface from this technique, called hyperspectral imaging, the founders hope to better understand the planet. To start, they plan to plot out clusters of plastic polluting the ocean, and to analyze forests’ moisture and chlorophyll levels to assess fire risk – and perhaps prevent the blazes that engulf more of the Earth with each passing summer.

The barely five-year journey from co-working space to outer space, however, is not nearly as quick as the journey that the data collected with Metaspectral’s technology from the ISS are expected to take: just 15 minutes to beam back to Earth.

Some of the technologies embedded in the services that Mr. Doumet and Mr. Tissera sell to clients have been around for a while. But their selling point, they hope, is how Metaspectral packages all its specialties. With a mix of computer-vision software, slick sensors and an open-source data-compression algorithm developed by space agencies, Metaspectral boasts that it can collect and analyze image data at extremely granular levels and send them where they need to go as quickly as possible.

That 15-minute data transmission from outer space effectively turns into real-time transmission on Earth, the Metaspectral founders say. And they’re already rolling out a wide range of uses, from forest-fire prevention to plastic-bottle identification on recycling lines.

In very basic terms, the artificial-intelligence branch of computer vision seeks out patterns – to tell that a cat is a cat because of its shape, or to identify other objects by colour. Mr. Tissera, Metaspectral’s chief technology officer, said the company envisions tapping into a vast, underserved market by expanding upon the definition of vision.

“We actually see what humans can’t see,” he said in a recent interview at the downtown Vancouver co-working space in which Metaspectral was first launched. Across the entire spectrum of light, including beyond what’s visible, every material has a unique “spectral signature” that can identify its composition with great specificity.

“We look deep into the infrared, and because we get so much more information, we’re actually doing scientific spectral analysis,” Mr. Tissera said.

Or, as chief executive officer Mr. Doumet put it: “This is computer vision on steroids.”

The commercial use cases for hyperspectral imaging are growing at a rapid pace, in large part because the hardware has become smaller, and champions such as Metaspectral are finding more efficient ways to transmit their data.

“This is a technology that I expect will go through a major transformation in the next three to seven years,” said Debra Laefer, a professor of urban informatics at New York University who uses hyperspectral imaging to study the composition of building and construction materials.

Prof. Laefer likens hyperspectral’s rise to that of light detection and ranging (LiDAR) technology, the cost of which has fallen greatly in recent years with increasing use in autonomous-vehicle testing and now smartphones. With broader adoption, “you get this economy of scale,” she said.

The ISS contract is a collaboration with HySpeed Computing, a Miami remote-sensor company that also specializes in hyperspectral image processing and analysis, using statistical and traditional machine-learning techniques. Ryan Reeves, director of science and technology for the ISS National Lab, said in an e-mail that such tech has the power to help humanity better understand Earth’s geography, vegetation, agriculture, water and other resources.

“Historically, real-time data generation and transmission have been a challenge on board the space station,” Dr. Reeves said. “This technology could lead to the ability to iterate on new scientific research and collaborate with researchers on Earth more rapidly, with the goal of accelerating progress on new discoveries and innovations that can ultimately benefit humanity.”

Mr. Doumet and Mr. Tissera came upon Metaspectral’s business model rather indirectly. Mr. Tissera was living in Australia late last decade, trying to watch UFC matches, but growing frustrated at the streaming quality. On one routine trip back to Vancouver in 2018, he encountered Mr. Doumet at the co-working space and pitched him on a video-compression algorithm that would compress and stream data in ways that would be less interruptive to, say, UFC streams.

After building a prototype compression algorithm, they started getting questions about managing and compressing hyperspectral data. “At that moment, we didn’t really have any idea what hyperspectral data was,” Mr. Tissera said. But when they learned that hyperspectral cameras can capture a truly unwieldy gigabit of data per second, the two saw a significant business case.

As other clients came forward, they realized new potential applications. Recyclers could identify specific types and brands of plastic bottles to sort similar plastics together with greater purity. Militaries could more clearly identify explosives or camouflage. Governments could analyze forests for moisture to see if a fire was about to break out. The company pivoted from its initial business model and became Metaspectral in 2020.

Despite patenting their own data-compression algorithm, Metaspectral’s staff have since decided to use an open-source one developed by space authorities but to implement it in their own proprietary way.

They’ve since built up a number of partners and clients on the ground and in the sky, and are hoping to set up office in the Vancouver area soon – after giving theirs up during the pandemic and returning to the co-working space where they first formed. Mr. Doumet and Mr. Tissera are also planning to raise a fresh round of financing after surviving on an angel round and customer revenue for the past few years.

And as they explore different industries to which they can apply hyperspectral imaging, the co-founders still have a very big space mission in front of them.

They’ve already impressed some space-tech peers. “They can actually do a high level of compression without losing any data – so we have a partner that really shines,” said Soren Pedersen, who works regularly with Metaspectral as a software sales manager with the Swedish space-tech company Unibap AB. “I would like more Metaspectral offerings on board.”