Astronomers have developed a machine learning algorithm that can detect satellite streaks in images of the night sky with high accuracy. The model makes it easier to interpret the data and could make it possible to eliminate the streaks that are increasingly causing problems in astronomy.

The technology won’t solve the problem of “photobombing” observations by Internet communications satellites, but it could help reduce their impact on some telescope images. Researchers presented the work at the General Assembly of the International Astronomical Union (IAU) in Cape Town last month.

“Machine learning and artificial intelligence can help because if you have enough data, you can classify that a satellite looks like this,” says Siegfried Eggl, an astrophysicist at the University of Illinois Urbana-Champaign. But satellite launches and developments are progressing at a “breaking pace,” he adds, and researchers are “doing our best to catch up.”

Growing threat

Over the past five years, companies such as SpaceX in Hawthorne, California, Eutelsat OneWeb in London and Amazon’s Project Kuiper in Redmond, Washington, have launched thousands of communications satellites into low-Earth orbit. Many more are planned, including a 12,000-satellite mega-constellation called G60 Starlink to be launched by Shanghai Spacecom Satellite Technology in China. “There are now about a million satellites on the register of ambitions for the future,” said Richard Green, director of the IAU Center for Protecting Dark and Quiet Skies from Interference from Satellite Constellations, in a session at the IAU General Assembly.

Astronomers urge global debate over huge swarms of satellites

These satellites provide high-speed broadband internet access to people around the world, but they are increasingly posing a nuisance to astronomers – appearing as bright streaks in sky images and can interfere with observations across the electromagnetic spectrum. Sensitive telescopes with large fields of view bear the brunt of this satellite pollution. The upcoming Vera Rubin Telescope, for example, could see more than a third of its images compromised, according to an estimate presented at the meeting.

“Astronomy is now big data science and there is no human being who can look at all the images taken every night and see the streaks,” says Eggl. “Machine learning can help here.”

To develop a program to identify satellite tracks in telescope images, María Romero-Colmenares, a data scientist at the University of Atacama in Chile, trained a supervised machine learning algorithm on tens of thousands of images captured by a network of telescopes across Chile. Spain, Mexico, Vietnam and South Korea. “We knew at what time and at what position (in the sky) we had to observe the satellite, and we did one observation with and one without the satellite,” says Romero-Colmenares, producing equal amounts of clear and contaminated images. When she and her colleagues applied the model to publicly available data from the WASP (Wide Angle Search for Planets) and Hungarian Automated Telescope Network projects, the algorithm was able to identify 96% of the satellite streaks.

“Unsustainable”: How satellite swarms pose a growing threat to astronomy

Detecting the streaks is an important step toward eliminating them from images and data, says Jeremy Tregloan-Reed, an astrophysicist at Atacama University who worked with Romero-Colmenares on the project. The next challenge will be to develop tools that can actually remove the satellite tracks while preserving the underlying data. This is only possible in cases where the satellite is not so bright that it saturates the pixels of an image and blends into the surrounding pixels, says Tregloan-Reed. If bleeding occurs, the underlying data cannot be saved.

By the end of next year, researchers hope to develop an open-source app and program that will allow observatories and amateur astronomers to identify and clean up contaminated images and data. Such measures are most likely to occur with small telescopes that have cameras with low sensitivity.

Star-like lightning

Other forms of satellite pollution are proving even more difficult to combat. When solar panels and other flat surfaces on satellites capture the light, they produce flashes that resemble short-lived astronomical transients, bursts of energy that can last from milliseconds to years.

“Since these flashes are very short, sometimes up to a millisecond, the movement of the satellite is negligible and we get a completely star-like flash,” says Sergey Karpov, astronomer at the Central European Institute of Cosmology and Fundamental Physics in Prague. There is “no real way to distinguish these flashes from astrophysical transients that we would like to detect – other than directly comparing their location with catalogs of satellite orbits,” he adds.

SpaceX tests black satellite to reduce threat to astronomy from ‘mega constellations’

Electronic equipment in satellites can also unintentionally emit radiation that interferes with observation of the afterglow of the Big Bang, says Eggl. Astronomers hope that studying this radiation, known as the cosmic microwave background, will provide answers to questions about the expansion of the universe. SpaceX’s next-generation satellites, which the company began launching last year, emit around 30 times more radiation than the previous generation. This type of radiation is unregulated and could affect entire observation bandwidths¹.

Eggl points out that AI tools can’t actually recover lost data and that the problem will get worse as more satellites are launched. “If you put white paint over it Mona LisaAt some point there’s nothing you can do, even if you train a machine learning algorithm on all of da Vinci’s paintings,” says Eggl. “They may guess what the painting might look like, but they can never reconstruct the lost data.”