Active compensation for perturbed coaxial reflecting space telescope using defocus point spread function and convolutional neural network

Active compensation for perturbed coaxial reflecting space telescope (CRST) is an effective and important means to improve the image quality, which has been widely used in the field of astronomical observation. The existing compensation methods have disadvantages of complicated structure, high-cost and poor efficiency. Aiming to solve these problems, in this paper, a novel active compensation method for perturbed CRST based on defocus point spread function (PSF) and well-trained convolutional neural network (CNN) is proposed. First, the optimal compensation strategy for perturbed CRST by adjusting the secondary mirror (SM) is presented, and the nonlinear mapping between the defocus PSF at the field point of interest and five kinds of misalignments of SM is established by utilizing CNN. Then, two alignment schemes either using single-field or multi-field defocus PSFs to obtain the adjustments of SM are proposed, and the simulation proofs show that the image quality can be improved by using our method to implement the active compensation for perturbed CRST. Finally, the influences of image noise and high-order figure errors of primary mirror (PM) on the compensation effect are investigated respectively. Compared with the existing compensation methods, the proposed method requires neither a dedicated wavefront sensor nor driving active correction components repeatedly nor a lot of iterations to optimize a metric function, which is suitable for in-orbit active fast compensation for perturbed CRST.