Satellites interfere with astronomical data - can AI offer a solution?
Satellites interfere with astronomical data - can AI offer a solution?
astronomers have developed a mechanical learning algorithm that can recognize satellite traces in pictures of the night sky with high accuracy. This model facilitates the data interpretation and could enable the stripes to be removed, which are increasingly causing problems in astronomy.
The technology becomes the Problem of "FotoBombs" of Internet communication satellites could not be solved, but could help reduce their effects on some telescopic images. Researchers praised the work last month at the general meeting of the International Astronomical Union (IAU) in Cape Town.
"Machine learning and artificial intelligence can help, because if you have enough data, you can classify, okay, that's what a satellite looks like," says Siegfried Eggl, astrophysicist at the University of Illinois Urbana-Champaign. But the number of satellite starts and developments happens at "rapid pace", he adds, and the researchers "do their best to catch up".
growing threat
In the past five years, companies such as SpaceX in Hawthorne, California, Eutelsat OneWeb in London and Amazon's Project Kuiper in Redmond, Washington, launched thousands of communication satellites into a low orbit. There are many more planned, including a 12,000 satellite megaconstellation called G60 Starlink, which is to be launched by Shanghai SpaceCom Satellite Technology in China. "There is now about a million satellites in the register of ambitions for the future," said Richard Green, director of the IAU center to protect the dark and calm sky against satellite constellation interferences, during a session at the IAU general meeting.
These satellites offer people's fast broadband internet access, but are increasingly disruptive for astronomers -They appear as light stripes in heavenly images and can influence observations over the entire electromagnetic spectrum. Sensitive telescopes with wide field of vision are particularly affected by this satellite contamination. For example, the upcoming Vera Rubin telescope could be estimated to see more than a third of his pictures.
"Astronomy is now science with large amounts of data, and there is no person who can watch all the pictures that are recorded every night and recognize the stripes," says Eggl. "Machine learning can help here."
In order to develop a program for the identification of satellite traces in telescopic images, María Romero-Colmenares, a data scientist at the University of Atacama in Chile, trained a monitored machine learning algorithm on tens of thousands of images that were recorded by a network of telescopes in Chile, Spain, Mexico, Vietnam and South Korea. "We knew when and where [in the sky] we should watch the satellite and made an observation with a satellite and one without", said Romero-Colmenares and created the same number of clear and contaminated images. When she and your colleagues used the model to publicly available data from the WASP (Wide Angle Search Search for Planets) and the Hungarian automated telescope network, the algorithm was able to identify 96 % of the satellite traces.
The detection of the strip is an important step towards elimination of these from pictures and data, says Jeremy Tregloan-Reed, an astrophysicist at the University of Atacama who worked on the project with Romero-Colmenares. The next challenge will be to develop tools that can actually remove the satellite traces while the data underneath is retained. This is only possible in cases where the satellite is not so bright that it is saturated by the pixels of an image and passes into surrounding pixels, says Tregloan-Reed. If there is overflowing, the underlying data cannot be saved.
The researchers hope to develop an open source app and a program by the end of next year that enables observator and amateur astronomers to identify and clean up contaminated images and data. Such measures are expected to be most successful in small telescopes with cameras with little sensitivity.star -like flashes
Other forms of satellite contamination prove to be even more difficult to manage. When solar modules and other flat surfaces capture the light on satellites, they create flashes, the Bureaous astronomical transients resemble energy, which can take from milliseconds to years.
"Since these flashes are very short, sometimes up to one millisecond, the satellite movement is negligible and we get a perfect 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 recognize - apart from comparing their position directly with catalogs from satellite tracks," he adds.
The electronic equipment in satellites can also emit unintentional radiation that disturbs the observations of the glow of the big bang, says Eggl. Astronomers hope that the study of this radiation, known as Cosmic microwave backlight , answer questions about the expansion of the universe . SpaceX’s the next generation of satellites that started the company last year radiates about 30 times more radiation than the previous generation. This type of radiation is not regulated and could endanger entire observation tapes.
Eggl points out that AI tools cannot really reconstruct lost data and the more satellites will be started the more satellites. "If you paint over the Mona Lisa with white color, there is nothing that you can do, even if you train a machine learning algorithm on all works from Da Vinci," says Eggl. "You may guess what the painting could look like, but you can never reconstruct the data you lose."
-
Bassa, C. G. et al. Astron. Astrophys. 689, L10 (2024).