Articial neural networks for centroiding elongated spots in Shack-Hartmann wavefront sensors

Abstract

The use of adaptive optics in extremely large telescopes brings new challenges, one of which is the treatment of images from Shack–Hartmann wavefront sensors. When using this type of sensor in conjunction with laser guide stars to sample the pupils of telescopes with diameters of more than 30 m, it is necessary to compute the centroid of elongated spots, whose elongation angle and aspect ratio are changing across the telescope pupil. Existing techniques, such as the matched filter technique, have been considered as the best ways to compute the centroid of elongated spots, but these are not good at coping with the effect of a variation in the sodium profile. In this paper, we propose a new technique using artificial neural networks. This technique takes advantage of the neural network’s ability to cope with changing conditions, and it outperforms existing techniques in this context. We have developed comprehensive simulations to explore this technique and we compare it with existing algorithms.