In the adaptive optics system of large-aperture ground-based telescopes, the wavefront sensor plays a crucial role. Pyramid wavefront sensors are increasingly favored by an expanding number of world-class telescopes. However, traditional wavefront reconstruction algorithms with pyramid wavefront sensors have limited ability to fit nonlinearity, resulting in restricted improvement in reconstruction accuracy. The kernel of deep learning lies in the ability of artificial neural networks to approximate nonlinear functions with arbitrary precision, which is well-suited for solving the nonlinear wavefront reconstruction problem of pyramid wavefront sensors and achieving more accurate wavefront sensing. This paper introduces the application of deep learning in pyramid wavefront sensors and Shack-Hartmann wavefront sensors, conducts a comparative analysis between them, and discusses potential future research directions.
|