Artificial Intelligence (AI) helps scientists create the most detailed map of the Moon.
Recently AI, who studied images of Selena, to understand how the craters look, discovered thousands of new objects of this kind on the surface of the satellite.
Recall that the lunar crater is a cup-shaped depression in the surface of the Moon, which has a relatively flat bottom and is surrounded by a ring-shaped raised shaft. It is believed that the absolute majority of such objects are impact type craters.
Traditionally, people counted and observed lunar craters during a visual survey of the satellite, but this method has its drawbacks (speed, for example).
“When it comes to counting craters on the moon, it’s a pretty archaic way, usually we need to carefully review the images, find and count the craters, and then calculate how large they are, based on the size of the image,” says one of the authors of the study, Mohammed Ali -Dib ( Mohamad Ali-Dib ), an astrophysicist from the University of Toronto.
Now scientists have created a new method for measuring the size and location of impact craters on the Moon. A new way will save both time and effort.
AI was taught to classify images and identify craters with a diameter of more than five kilometers
The new algorithm is an artificial neural network that tries to mimic the way that the brain processes information. First, the AI was trained with images of the Lunar Reconnaissance Orbiter , covering about a third of the Selena surface, then the program was shown with another third of the lunar landscape.
After that, scientists tested the trained neural network on the remaining third of the Moon. In all, there were about 90 thousand pictures.
The algorithm was also trained to determine the edges of the craters, which were then checked at the base of the detected craters. He used this information to confirm that the shape of new objects corresponds to known forms of the crater. In turn, this made it possible to distinguish craters from other objects.
AI revealed 92% of the known craters in the studied region, but, most interesting, he managed to notice another 6883 new. The technology worked so well that it could determine twice as many craters as traditional manual counting.
It is noted that the technology is based on a convolutional neural network, it is a class of machine learning algorithms that is successfully used for computer vision, robots and self- controlled machines.
Previous computer algorithms designed to more accurately and quickly count the number of craters, have not been as good in the detection of craters in the regions that have not studied algorithms in training, says astrophysicist Ari Silburt ( to Ari Silburt ) from the University of Pennsylvania.
According to him, as soon as they improve their model a little, scientists will be able to use it to detect hundreds of thousands of currently unidentified craters with a diameter of less than five kilometers.
It is noteworthy that the new AI is able to distinguish craters from the images of Mercury.
According to experts, the program can be used for the cataloging of tracks, which has arisen as a result of strikes, and on other satellites and planets. Among them, for example, Mars, the asteroid Vesta, the dwarf planet Ceres and the icy satellites of Jupiter and Saturn.
By the way, experts say that thanks to AI on the satellite, you can determine the best place for the future lunar base of the colonizers.
The results of the study are presented on the website of the preprints arXiv.org.
We add that the authors of the project “Vesti.Nauka” (nauka.vesti.ru) told us about other possibilities of artificial intelligence: in particular, it will begin to build maps of deposits instead of geologists and discover new planets. Moreover, AI has been taught to accurately predict earthquakes.