Liquid crystal wavefront correction based on improved machine learning for free-space optical communication

Abstract

In order to suppress the impact of atmosphere turbulence on the space laser communication link, the wavefront correction technology of a liquid crystal spatial light modulator (LCSLM) is studied. Combining with the control mode of the LCSLM, we propose an improved deep learning approach that restores the input image features into the wavefront and then controls the LCSLM to compensate for the phase distortion. This method does not have Zernike coefficient truncation and does not require the calculation of coefficient matrices, thus improving the accuracy and efficiency of the algorithm. At the same time, as for its powerful phase fitting ability, the LCSLM can be used as a turbulence simulator to construct datasets. During the training process of the neural networks, a calibration between the LCSLM and deep learning is established. Finally, a spatial optical coupling experimental system is built. The results show that, under different atmospheric conditions, the liquid crystal wavefront correction method has a significant improvement in terminal coupling efficiency and has certain application prospects in the field of free-space optical communication.