Artificial intelligence only takes 5 seconds to detect a new crater on Mars

While NASA scientists need up to 40 minutes, artificial intelligence can analyze a picture of the Martian surface in just 5 seconds.

Currently, to identify newly formed craters on Mars, scientists will have to spend about 40 minutes analyzing a picture of the Red Planet's surface taken by the camera on the Reconnaissance Orbiter. Mars (MRO) by NASA.

Meteor craters on Mars.

If a scientist detects signs of a crater in those images, that must be confirmed using a higher-resolution photo taken by another MRO instrument.

This method of detecting new craters on Mars makes it easy to determine approximate dates for when they formed. For example, if a crater isn't pictured from April 2016 but is in a photo from June 2018, scientists know it must have formed around that time.

By studying the characteristics of craters for which they know their age, scientists can then estimate the age of larger craters.

This information could improve scientists' understanding of the history of Mars and help plan new missions to the Red Planet.

The MRO has been imaging the Red Planet's surface for 15 years, and in that time it has taken 112,000 lower-resolution images, with each image covering hundreds of miles of Mars' surface.

To relieve scientists of the burden of manually analyzing all those images, the researchers trained an algorithm to scan similar images for signs of new craters on the planet. this planet. The problem is worth mentioning, and it only takes about 5 seconds for each photo.

Besides, to train the artificial intelligence (AI) to analyze their images to detect new craters on Mars, the researchers started by feeding it nearly 7,000 images from camera captures. Some of the prominent new craters were confirmed by the High Resolution Imaging Experience (HiRISE) orbital camera, while others were not.

AI could help scientists identify more craters on Mars.

After training, the next step is to allow the algorithm to analyze all the images captured by the camera. To speed things up, the researchers ran the AI ​​on a cluster of supercomputers at NASA's Jet Propulsion Laboratory (JPL).

"It would be impossible to process more than 112,000 images in a reasonable amount of time without distributing the work across multiple computers. The strategy is to break the problem down into smaller chunks that can be solved in parallel," the researchers said. scientists said.

With the power of all those computers combined, the AI ​​can scan an image in just 5 seconds. If it detects something that looks like a new crater, NASA scientists can test it for themselves using HiRISE.

NASA confirmed that AI has discovered the first new crater on Mars, and so far, it has helped scientists detect dozens of new impacts.

"The data is always there, it's just that we haven't seen it ourselves," says JPL computer scientist Kiri Wagstaf .

In the future, AI could help scientists identify more craters on Mars, potentially within weeks of their formation, or even craters on other planets.

Ingrid Daubar, a planetary scientist who helped create AI, added: " The ability to use machine learning to really drill down into large data sets and find things that we wouldn't otherwise find. It's really exciting. This is just the beginning. We look forward to finding more in the future."