In 2019, 11,482 Hubble Asteroid Hunter volunteers helped train an AI that scanned 37,000 Hubble images; the combined effort uncovered 1,701 asteroid trails, including 1,031 unknown objects

Major discoveries often begin with an accident. In 2019, a group of astronomers turned to ordinary citizens to help sort through old Hubble Space Telescope images, looking for asteroids that photobombed pictures meant for something else entirely. According to a 2024 study published in Astronomy & Astrophysics by astronomer Pablo García Martín and his team, that effort eventually found 1,701 asteroid trails, 1,031 of them objects never before recorded.

A photobombing asteroid gave the game away

Asteroids range in size from tiny pebbles to large rocks. Cataloging them is really difficult, the study explains, because they are faint and they keep moving in their orbits around the Sun. You can't just point a telescope at fixed coordinates and expect to find one there. Hubble was never designed to hunt asteroids anyway. Its usual targets are distant galaxies and stars, but one famous picture of the galaxy UGC 12158 features a curving white line that crosses the picture. Researchers note that the streak is a nearby asteroid that floated through Hubble’s sight while it was busy taking pictures of the galaxy. The curve appears because Hubble is in orbit around Earth during observations, and that changing viewpoint helped scientists determine the asteroid's distance.


Recruiting more than 11,000 citizen scientists

It was not feasible for a small research team to manually sift through Hubble's whole picture archive. So in 2019, researchers at the European Space Agency’s technology and science centers joined forces with Zooniverse, a popular citizen-science platform, and Google to launch the Hubble Asteroid Hunter project. According to the official statement on the project by the European Space Agency, 11,482 volunteers from all over the world searched through 37,000 images taken by Hubble over a period of 19 years, and together they recorded almost two million individual marks.

Turning volunteer clicks into an AI model

Human eyes are good at spotting patterns, but they can't go on checking each and every image indefinitely. An earlier paper on the project by Sandor Kruk and colleagues reports that the markings made by volunteers were used for training the machine-learning model. Once trained, the model was able to process far more images than any human group would ever be able to process alone. The study suggests that this combination of citizen contribution and machine learning could also be useful for other image databases as well.

What the search really found

Between the citizen-science project and the machine-learning search, the team identified 1,701 asteroid trails. Among these, 670 could be matched to asteroids previously listed, and 1,031 were not matched to any entry in the official Solar System database, indicating that these were probably new discoveries. As the European Space Agency's press release on the discovery explains, around 400 out of the newly discovered objects have diameters smaller than a kilometer. "We were surprised with seeing such a large number of candidate objects," García Martín said in that same release, adding that the result confirms a smaller population of main-belt asteroids scientists had only suspected existed until now.

What the tiny asteroids reveal about the solar system's past

These asteroids are found mostly in the main asteroid belt, a region of rocky bodies between Mars and Jupiter. The size of each asteroid was calculated based on how bright it looked in comparison to the distance from which it appeared. It also supports one of the scientific theories that have been debated for a very long time by scientists: that many of these small asteroids have come from much bigger asteroids due to breakups in collisions over millions of years, just like broken pottery pieces, a finding that challenges the competing theory that these small fragments simply formed that way from the start. The research group plans to do further research on these asteroids, as well as characteristics such as size and orbit. However, the study highlights that it would not be possible to calculate their full orbits because many of these trails were taken by the Hubble Telescope long ago and cannot be observed again in the same way.

Why you should care about this

You don’t need a telescope or a degree in physics to help real scientists make discoveries. The Hubble Asteroid Hunter project shows how thousands of people, who dedicated only a few casual hours working at home on the computer, could make discoveries, which otherwise would go unnoticed both by computers and scientists working on their own. NASA still offers similar citizen-science projects that anyone anywhere in the world can join. Next time an old photo throws up something unexpected, it could be the start of a new discovery.