Improving the Accuracy of the Adaptive Optical Wavefront Correction System of High-Power Lasers Based on Convolutional Neural Networks

The paper examines the possibility of increasing the accuracy and speed of calibration of an adaptive wavefront correction system for high-power laser systems based on convolutional neural networks. The work is based on solving the inverse problem of reconstructing the profile of phase distortions of the wavefront from two images captured in the vicinity of the waist of the focusing system. The calibration speed has been shown to increase by more than 10 times compared to classical methods for calibrating an adaptive optical system.