Chinese astronomers have identified more than 109,000 previously unknown craters in the Moon's low- and mid-latitude regions using artificial intelligence trained on data from Chinese lunar orbiters.
Lunar impact craters, created by meteor strikes, serve as "fossils" preserving the solar system's history. Varying in size, shape, and often overlapping or eroded, their detection has long been labor-intensive and complex.
To address this, researchers at Jilin University developed a deep neural network—a sophisticated machine learning model using interconnected layers of calculations—trained on thousands of known craters. Once trained, the AI efficiently detected new ones.
Analyzing data from the Chang'e-1 and Chang'e-2 orbiters, the AI revealed 109,956 additional craters, increasing the known total in these regions by more than tenfold.
"This is the largest lunar crater database with automatic extraction for the Moon's mid- and low-latitude regions," said lead author Chen Yang from Jilin University.
Most are small to medium-sized (1 to 100 km in diameter), per the study in Nature Communications. Some are larger, irregularly shaped, highly eroded, and up to 550 km across.
The algorithm also dated nearly 19,000 craters, spanning the Moon's five geological periods, with some tracing back about four billion years.
Next, the team plans to incorporate Chang'e-5 lander data—the first lunar samples returned to Earth in over 40 years—and extend the method to other solar system bodies like Mars.
