Researchers from the Department of Systems Innovation, School of Engineering, The University of Tokyo successfully developed an innovative and efficient AI-based automatic rock identification algorithm. Rock particles ubiquitously exist on both terrestrial and extraterrestrial solid surfaces throughout the solar system. A detailed understanding of the characteristics of rocks, including their size, shape, orientation, and spatial distribution, is essential for both science and engineering purposes. While rock mapping may appear to be a simple and easy task, it is extremely challenging due to the generally huge number of rocks, ambiguous outlines, and overlapping particles existing on planetary and small-body surfaces. As a result, conventional manual analysis often yields unreliable and inconsistent results, making science interpretation and reproducible analysis difficult to achieve.
To address this problem, project researcher Yuta Shimizu, a pioneer in the field of data-driven geology and planetary science at the University of Tokyo, developed a new technology that automatically identifies thousands of rocks within a few seconds. With his two colleagues, he applied this algorithm to the images from the asteroids Ryugu and Bennu, provided by Hayabusa2 (JAXA) and OSIRIS-REx (NASA) missions, respectively, and revealed their distinct evolutionary histories due to their different rotation rates.
“We have successfully developed a robust, fast, and reliable new algorithm for identifying millions of rocks.”, Shimizu said. “We identified more than 3.5 million rocks in total on the images of asteroids Ryugu and Bennu and found that even small variations in rotation period, on the order of a few hours, can drive complex and diverse evolutionary pathways of asteroids in the solar system.”
Patrick Michel, Global Fellow at the University of Tokyo and Director of Research at CNRS (French Scientific Research National Center) anticipate this algorithm will have immense potential for future space exploration missions. “This new AI identification technique is highly efficient and will provide valuable support for our space explorations of asteroids.” Patrick said. “We are living in the golden age of small-body exploration, with numerous missions already in flight or under development by various space agencies around the world, serving both scientific research and planetary defense. This advancement could greatly benefit these efforts, for the best of international cooperation.”.
Accurate identification of rocks and their physical properties is crucial in a variety of industries, directly influencing decisions in mining, civil engineering, and disaster prevention. This algorithm offers a new solution for these fields.
“Our new algorithm has the potential to completely change the way rocks are analyzed in a variety of industries”, Shimizu said. “The technology in this work has broad potential not only for planetary science, but also for industrial applications, from continuous slope monitoring for disaster mitigation to real-time material management in mining and construction, as well as rapid urban infrastructure inspection and agricultural soil analysis.”
Papers
Journal: Scientific Reports
Title: Diverse evolutionary pathways of spheroidal asteroids driven by rotation rate
Authors: Shimizu Yuta, Miyamoto Hideaki*, Michel Patrick
DOI: 10.1038/s41598-025-94574-1