A new method of scanning the Moon has appeared, which allows you to automatically classify the features of the Earth’s satellite using telescope images. This will increase the efficiency of site selection for research.

Choosing a place to land or explore the moon is a complex process, the scientists noted. The visible area of ​​the lunar surface is larger than the territory of Russia, it consists of thousands of craters. The choice of future sites for planting and research can be reduced to the choice of the most promising sites for construction, minerals or potential energy resources. However, eye scanning such a large area looking for objects perhaps several hundred meters in size is laborious and often inaccurate.

Scientists at the Chinese University of Hong Kong have now applied machine learning and artificial intelligence (AI) to automate the identification of promising landing areas and lunar exploration.

“We are looking for lunar features such as craters that are sources of energy resources such as uranium and helium-3, a promising resource for nuclear fusion,” the team noted. “Both have been found in lunar craters and could be a useful resource for replenishing spacecraft fuel.”

The next task was to develop a computational algorithm that can be used to estimate craters and territories on the Moon. “We solved this problem by creating a deep learning system called high-resolution-moon-net. It consists of two independent networks with the same network architecture to simultaneously detect craters and study the surface, ”the researchers noted.

The team’s algorithm achieved an accuracy of 83.7%, which is much higher than other methods.