A deep-learning artificial intelligence technology has discovered over 6,000 new craters on Earth's moon. These never-before-seen craters were mapped by an AI tool in just several hours.
Researchers are using neural networks to study and map the lunar surface. Using a new technique, they have successfully counted new pockmarks on the moon — some 6,000 of them — through available datasets from previous lunar observation information.
The moon is dotted with vast number of craters some billions of years old.
"Basically we need to manually look at an image, locate and count the craters and then calculate how large they are based off the size of the image," says Mohamad Ali-Dib, from the Centre for Planetary Sciences at University of Toronto Scarborough.
Deep-Learning Technique
The researchers first trained the convolutional neural network on a dataset covering two-thirds of the moon. They then tested the neural network on the remaining third of the moon. The results yield 92 percent of craters from human-generated test sets and almost twice of the total number of crater detections.
They were able to spot about 6,000 previously unidentified craters on the moon's surface.
Out of the new craters discovered, 15 percent are smaller in diameter than the minimum crater size in the ground-truth dataset. The errors compared to the human-generated datasets are only 11 percent or less, making the deep-learning tool useful in automatically extracting craters information on various solar system bodies.
The deep learning framework utilized for the research is an NVIDIA Tesla P100 GPU and the cuDNN-accelerated TensorFlow.
The same network also successfully detected craters on Mercury, which has a completely distinct surface compared to the moon.
Same Technology Used in Self-Driving Cars
As its name implies, convolutional deep neural network is comprised of one or more convolutional layers and followed by one or more fully connected layers in a standard multilayer neural network.
This network is designed to maximize the 2D structure of an input image. In the case of mapping of lunar craters, the algorithm analyzed the data taken from elevation maps gathered from orbiting satellites.
This learning algorithm has been used to power robots and self-driving cars.
Through deep learning, machines have the potential to identify even small craters and uncover more information on other bodies in the solar system including Mercury, the dwarf planet Ceres, the asteroid Vesta, or the icy moons of Jupiter and Saturn.
The technique was developed by Ali-Dib, research coauthor Ari Silburt of Penn State University, postdoctoral researcher Chenchong Charles Zhu, and a group of researchers at the Centre for Planetary Sciences and the Canadian Institute for Theoretical Astrophysics.
A pre-published version of the study is under review in the journal Icarus.
