NASA is Using Artificial Intelligence (AI) to Detect New Craters on Mars
기사입력 2021.02.05
NASA's new AI tool is helping the ongoing investigation of Mars and its environment.
  • It is has been almost 15 years since NASA launched its Mars Reconnaissance Orbiter (MRO) into space. Since then, the orbiter has returned 404 Terabits (TB) of data, and counting. The main function of the spacecraft is to collect and relay any geographical and weather data about Mars.
  • Mars Reconnaissance Orbiter (MRO)/ Image credit to NASA Science Mars Exploration Program
    ▲ Mars Reconnaissance Orbiter (MRO)/ Image credit to NASA Science Mars Exploration Program

    Mars Reconnaissance Orbiter (MRO) is equipped with at least ten scientific instruments such as High-Resolution Imaging Science Experiment Camera (HiRISE), Context Camera (CTX), and Compact Reconnaissance Imaging Spectrometer (CRISM). Using HiRISE, a team of scientists and AI researchers at NASA’s Jet Propulsion Laboratory (JPL) in Southern California, were able to develop a machine-learning tool for HiRISE’s camera to be able to discover craters on Mars, even small ones which are about 4 meters (13 feet) in diameter.

    Before the machine-learning tool, NASA relied on people and MRO’s Context Camera, which takes low-resolution images, to discover craters on Mars. However, HiRISE’s camera takes high-resolution images that are able to even view tracks left by NASA’s Curiosity Rover, a rover that travels around Mars to learn about its environment.

  • AI machine learning spots a crater on Mars/ Image credit to NASA Science Mars Exploration Program
    ▲ AI machine learning spots a crater on Mars/ Image credit to NASA Science Mars Exploration Program

    Normally, it takes about 40 minutes or more for researchers to scan one Context Camera image. However, the AI-powered tool called automated fresh impact crater classifier is able to scan an image in about five seconds. To train the classifier tool, researchers initially fed the program 6,830 Context Camera images of previously discovered impact craters. Then, the tool was fed images of no impacts, so it learns what not to look for. Once the classifier tool was trained, NASA then fed all of Context Camera’s 112,000 images to look for meteor impacts on Mars for analysis.

    The classifier tool is far from perfect as it still requires a human to check and confirm image findings. However, the tool drastically cuts down time, money, and manpower. Currently, the AI-tool is being used on Earth-bound computers, but the hope for NASA is to be able to run the tool onboard in future orbiters. Once that happens, then the goal is to be able to learn more about Mars, meteor impacts, planetary orbit, and the solar system.