Space imagery (referring here to image data used in space-related research and applications, whether acquired in space or on Earth) has recently become a focus of the Image Processing and Pattern Recognition groupe. The mutliple challenges posed by this domain, such as detecting low-contrast or morphologically specific objects (such as stars or cosmic particles), combined with the need for large-scale data processing, make it an ideal applicative field for the algorithms developed in the team. Our work is part of interdisciplinary collaborations with research institutes specializing in space sciences.

Our contributions in space imagery

Ongoing projects in space imagery within the Image Processing and Pattern Recognition group include:

• Meteor detection and trajectory reconstruction As part of the CABERNET project (CAmera for a BEtter Resolution NETwork), conducted in collaboration with the Laboratoire Temps-Espace (Observatoire de Paris - PSL), the team contributes to the automatic detection and identification of meteors in images continuously captured by ground-based stations. The CABERNET project endgoal is to reconstruct meteoroid orbits in 3D to better predict meteor showers and outbursts. Our work focuses on designing and improving detection and classification algorithms, leveraging both classical image processing techniques such as mathematical morphology, and modern machine learning approaches.
• Exploration of Cassini mission data The team also participates in the large-scale analysis of the image dataset from NASA’s Cassini mission, which orbited Saturn for over 13 years and captured several hundred thousand images. In collaboration with the Laboratoire Temps-Espace, we develop AI-based algorithms for the automatic classification of astronomical objects in these images (such as Saturn’s moons or cosmic rays) with the aim of deepening our understanding of the Saturnian system.
• Meteorite classification via electron microscopy In collaboration with the Institut de Minéralogie, de Physique des Matériaux et de Cosmochimie (Sorbonne Université) and the Muséum National d’Histoire Naturelle, the team develops image processing and classification algorithms for the semantic segmentation of mineral phases within chondrite meteorite samples acquired with scanning electron microscopy as well as the meta-classification of the meteorite based on its chemical and mineralogical composition, serving as a first crucial step for any further in-depth study of the formation of our solar system.