Hubble Asteroid Hunter project FEATURES
However , machines are not yet capable of making new discoveries on their own , as they lack the intuition , creativity , and distraction of the human brain .
That ’ s where citizen science comes in . By harnessing the collective effort of thousands of volunteers from across the world and the power of an artificial intelligence ( AI ) algorithm , we found over 1000 asteroids hiding in the ESA Hubble Space Telescope archives . The project was born after-crossmatching the orbits of known Solar System Objects ( asteroids , comets , trans-Neptunian objects , etc .) with archival observations from Hubble , Herschel , and XMM-Newton ( Racero et al . 2022 ). We found that some predicted asteroid positions had no object or had another moving object in the images . We decided to call citizen scientists to help and collaborated with the Zooniverse team , the largest platform for online citizen science projects , to build an asteroid hunting crowdsourcing project , Hubble Asteroid Hunter ( www . asteroidhunter . org ). We launched the project ahead of the International Asteroid Day , on 30 th June 2019 , asking volunteers to identify asteroids in 19 years of Hubble Space Telescope observations taken between 2002 and 2021 with the Advanced Camera for Surveys and the Wide Field Camera 3 instruments . The response of the public was overwhelming , with over 20,000 classifications per day during the first days .
Identifying asteroid trails in the images automatically turns out to be challenging without having a representative set of examples . The Hubble Space Telescope orbits around the Earth , while asteroids cross its field of view . Due to the motion of the spacecraft , asteroids appear as curved trails with a variety of shapes ( see two asteroid trails passing in front of a galaxy targetted by Hubble ). To solve this problem , over 11400 members of the public analysed and classified Hubble images during one year on www . asteroidhunter . org , identifying more than 1000 trails . This was an ideal training set for an automated algorithm based on AI . To be able to classify images automatically , we teamed up with Google and used the cloud-based AI algorithm , AutoML . With the use of AutoML were were able to classify 150,000 images ( corresponding to 37000 Hubble ‘ composite ’ or stacked images ) in just 7-hours . This was possible only with the initial help of citizen scientists . The combination of citizen science and AI resulted in a final dataset of 1701 asteroid trails ( a sample of these trails is shown in Figure 2 ). Roughly one third of these trails could be identified and attributed to known asteroids in the IAU ’ s Minor Planet Centre ( IAU Minor Planet Center ), which is the largest database of Solar System objects . This left 1031 unidentified trails that could be potential new asteroids , fainter than magnitude 22-23 , which would not easily identifiable in typical ground-based surveys .
▲ FIG . 1 : Two unidentified asteroids crossing paths in the foreground of dwarf galaxy AGC111977 . Credit : ESA / Hubble & NASA , J . Cannon ( Macalester College ), Kruk et al . 2022 ( source : * ESA - One galaxy , two asteroids : https :// www . esa . int / ESA _ Multimedia / Images / 2020 / 06 / One _ galaxy _ two _ asteroids ).
Although it ’ s not possible to track the orbits of the newly detected asteroids , as the Hubble images were taken many years ago , we can still use the telescope to determine the distance to them and constrain their orbits . This is through the so-called parallax effect , imprinted by the fast motion of Hubble around the Earth and the motion of the asteroid in the sky . Most of the unknown asteroids are likely located in the Main Asteroid belt , between the orbits of Mars and Jupiter , where most asteroids are situated . Knowing the distance , the observed brightness of the objects can be eventually translated into a physical size . These
▼ FIG . 2 : Mosaic of asteroid trails detected in the Hubble Asteroid Hunter citizen science project in different images from the NASA / ESA Hubble Space Telescope . Each of these images was assigned a colour based on the time sequence of exposures , such that the blue colours represent the first exposure in which the asteroid was captured , and the red colours represent the last . Credit : ESA / Hubble & NASA , M . Zamani ( ESA / Hubble ), Kruk et al . 2022 , source : Asteroid Trails Mosaic | ESA / Hubble ( esahubble . org )
EPN 54 / 2 29