Sometime between March 2010 and May 2012, a meteor streaked across the Martian sky, where it broke into pieces and crashed into the Red Planet’s surface. The resulting craters were relatively small at 13-feet in diameter. The smaller features are on the planet’s surface, the more difficult it is to discover the from the Mars orbiters.
For the first time, scientists have spotted craters from that meteor using artificial intelligence and machine learning. Planetary scientists say that using AI in this matter is a milestone. AI researchers from NASA’s Jet Propulsion Laboratory worked together to develop the machine-learning tool that discovered the impact craters. Researchers hope that the new AI will save time and increase the volume of findings.
The typical method used by scientists to discover craters such as these is to spend hours each day studying images captured by the Mars Reconnaissance Orbiter looking for surface phenomena. Scientists studying Mars have relied on Mars Reconnaissance Orbiter data to find over 1000 new craters over its 14 years in orbit.
Only the blast marks around the impact stand out with the individual craters unable to be seen in the image above. The next step in the process will be to look at the area using the High-Resolution Imaging Science Experiment known as HiRISE. That instrument is powerful enough to see details as fine as tracks left by the Curiosity Mars rover.
Searching manually through photos looking for surface phenomena requires 40 minutes or so for researchers to scan a single Context Camera image. To speed the process, researchers created a tool called the automatic fresh impact crater classifier. Training the classifier required researchers to feed it 6830 Context Camera images, including those locations with impacts previously confirmed using HiRISE. While it takes 40 minutes for a human to analyze an image, the AI tool needed only five seconds. Researchers point out that the classifier still requires a human to check its work despite all the computing power.