Vision Modelling in Astronomy : Recent Results 1 Artiicial Vision and Astronomical Images

Many researches were carried on the automated identiication of the astrophysical sources, and their relevant measurements. Large astronomical images may contain typically a large set of stars and diiuse objects. The automated analysis requires sequence of algorithms, for detecting, for identifying and for measuring these sources. Many vision models have been developed for this task. Their use depends on the image content : quite only stars, crowed elds, deep images with many faint galaxies, gaseous nebulae, .. In particular we have developed a multiscale vision model well suitable for analyzing complex structures such like interstellar clouds, galaxies, or cluster of galaxies. A large part of the today applications of astronomical image processing concerns the image analysis, i.e. the capability to extract source positions, magnitudes and pattern parameters. Image analysis is often done by an interactive process. This task is demanding, and then many researches were carried on the automated identiication of the astrophysical sources, and their relevant measurements. This purpose is obviously particularly interesting for large images, such as the ones obtained with Schmidt or other wide eld cameras. Large astronomical images may contain typically a large set of point-like sources (the stars), some quasi point-like objects (faint galaxies, double stars,..), some moving asteroids, and some complex and diiused structures (galaxies, gaseous nebulae, planetary stars, clusters, etc.). The image content depends on many conditions: The sky region to be analyzed. In galactic plane, we detect essentially stars and nebulous objects; The magnitude of the faintest objects which can be detected, deeper the observation, greater the proportion of faint galaxies compared to the stars. The automated analysis requires sequence of algorithms, for detecting, for identifying and for measuring the sources. We call as vision model the corresponding rules. The same model can be implemented by diierent softwares, with a similar analysis. Often the astronomers give reference to a peculiar package, so that a model is called by this package name. 2 The stellar eld model. At a rst glance, an astronomical image is made up a set of quite punctual spots due to the stars. For observations with modern electronic devices, such as CCD, each stellar proole is proportional to the Point Spread Function (PSF), i.e. the image of a punctual source seen 1