Asteroid Trails - Machine Learning Group
ESA’s Hubble Space Telescope Archive contains more than a hundred Tb of data. All these images were used in the past to conduct major breakthroughs in Astrophysics, but we still can find hidden treasures laying within them. We are using Machine Learning techniques to analyse images from ACS/WFC and WFC3/UVIS instruments looking for serendipitous asteroid trails passing between the telescope and its target. Starting from citizen science project Asteroid Hunter, we used data classified by volunteers to train Google’s AutoML Object Detection model to identify asteroid streaks.
Our goal: get relevant scientific information to characterise these objects and check whether they are already known by humankind.