Artificial Intelligence takes the field to ‘clean’ the images of the cosmos

Artificial Intelligence (AI) is ready to take the field alongside astronomers, to ‘clean’ the images of the universe obtained from ground-based telescopes: thanks to an algorithm already used for the photos and adapted by researchers led by the American Northwestern University and China’s Tsinghua University, AI is able to eliminate the disturbances caused by the Earth’s atmosphere, making photos much sharper and therefore helping researchers to obtain more accurate data. The new instrument, published in the Monthly Notices of the Royal Astronomical Society, has been freely made available to the scientific community.

When the light from distant stars, planets and galaxies reaches our atmosphere, it is distorted by the turbulent movements of the air: this is why the best ground-based telescopes are all located at high altitudes, where the atmosphere is thinner. This interference becomes a problem when it comes to extracting data from photos: for example, gravitational effects are detected by studying the shape of galaxies, but distorted images can lead to miscalculations.

To overcome this obstacle, researchers led by Emma Alexander of Northwestern University and Tianao Li of Tsinghua University combined a well-known photo-enhancing algorithm with an AI-based network trained on astronomical images. The new tool recorded errors that were 38.6% lower than classic image cleaning methods and 7.4% lower than more advanced methods.

Furthermore, the algorithm was also trained with images that simulate those that will be obtained from the Vera Rubin Observatory (Vro) under construction in Chile, which should come into operation in 2024: this means that, when the new telescope begins to observe the sky, the AI-powered instrument will be ready to help astronomers analyze the long-awaited data.