Artificial Intelligence and the Hubble Archive. 1000 unknown objects in the background!

Astronomers from the European Space Agency (ESA) have joined amateur scientists and artificial intelligence. Together they made a great discovery – More than 1,000 new asteroids have been identified in our solar system.

During the work, a new approach was used to document objects found in older datasets. The researchers point out that this method can also be applied to other data sets.

The research team was led by Pablo García Martín from the Department of Theoretical Physics at the Autonomous University of Madrid (AMU). Specialists point out that monitoring asteroids that come in various sizes and shapes is very difficult.

Researchers know Hubble's position because it takes pictures of space and then measures the curvature of the resulting images Arrogant planet. On its basis, it is possible to determine the distance of a particular object, as well as the shape of its orbit.

As Martin says: “We're seeing fewer and fewer main belt asteroids, deeper and deeper.” We were surprised by so many possible things. There have been some hints that this cluster exists, but we now confirm this with a random sample of the asteroid collection obtained using the entire Hubble archive. This is important for providing insight into evolutionary models of our solar system.

According to accepted theory Small asteroids were created by the collision of larger asteroids – This has been going on for billions of years, and the appearance of things is constantly changing. There is also a competing theory that states that such small celestial bodies now look almost identical to how they did at the time of their formation.

In 2019, the citizen science project was launched Hubble Asteroid Hunter (HAH), which aims to identify asteroids in archived Hubble data. In total, the entire team consists of 11,482 volunteers who have so far analyzed 37,000 Hubble images over 19 years.

In total, it was delivered approx Two million metadata. The next step was to use a special training set for an automatic asteroid recognition algorithm based on machine learning. It was the result of this undertaking Discovering 1031 previously unknown asteroidsOf which, approximately 400 are less than 1,000 metres.

– The positions of asteroids change over time, so they cannot be found by simply entering the coordinates, as they may be missing at different times. As astronomers, we don't have enough time to look at all the images of an asteroid. So we came up with the idea of ​​working with over 10,000 citizen science volunteers on the review Hubble's vast archives – Martin said.