Off-axis aberration reconstruction of liquid mirror telescope based on neural network

The inherent aberrations existing at the off-axis angle of liquid mirror telescopes (LMT) limit their application in astronomy and other areas. In this paper, a new method of wavefront reconstruction based on neural networks is used to correct the off-axis aberrations in LMT. Firstly, the components of the off-axis aberration including defocus, astigmatism and coma which have the greatest influence on the imaging results, are analyzed. Then the nonlinear relationship between the aberrations and the image is discussed. Finally, a model using the convolutional neural network (CNN) is established to fit the nonlinear relationship. The results of network training and verification show that the network can predict the original aberration information quickly and accurately.