Loïc Denis, MCF HDR, Hubert Curien laboratory, Image Science & Computer Vision team
The observation of extra-solar planetary systems (planets orbiting stars other than our sun) is a recent revolution in astronomy.
Almost 99% of exoplanets have been detected by indirect methods, analysing the motion of the host star in gravitational interaction with the exoplanet(s).
Direct imaging, on the other hand, seeks to produce an image of the environment close to the star and thus to detect and characterise exoplanets or dust discs typical of planetary systems in formation. This is a real technological challenge because of the very small angular separation between the star and its planetary system (requiring very high angular resolution) and the formidable contrast between the star and the exoplanets (in the infrared, a young Jupiter-type exoplanet is about a million times less bright than its sun, a large telluric planet is about a billion times less bright).
The two most powerful direct imaging systems currently available, the European SPHERE instrument and the American GPI instrument, combine a large telescope (8m in diameter), dynamic correction for the effects of atmospheric turbulence (adaptive optics) and a differential imaging observation strategy (acquisition of a sequence of images in which slight variations in the signal of interest facilitate its detection).
In collaboration with the Lyon observatory, a doctoral student and a teacher-researcher from Télécom Saint-Etienne have developed algorithms for detecting exoplanets with unequalled sensitivity. Their methods are based on recent data science techniques, in particular modelling and statistical learning of the spatial structure of signal fluctuations. The software developed has since been applied to analyse data collected during hundreds of nights of observation at the Very Large Telescope in Chile, allowing the reconstruction of the trajectory of exoplanets along their orbit or the characterisation of their spectrum.