The compensation of atmospheric turbulence for vortex beam based on deep learning

Beam propagation in free space will inevitably be disturbed by atmospheric turbulence (AT). In this paper, deep convolutional neural network (CNN) is used to extract the phase information of Gaussian beam. Through the training, the loss function value decreases from about 0.04 to less than 0.005, and the loss on the test set also decreases to about 0.005, and the phase information is extracted successfully. The phase information of Gaussian beam under turbulent conditions is extracted by CNN network to compensate the transmission of vortex beam under turbulent conditions. From the research, it is found that extracting turbulent phase through CNN network and compensating vortex beam with the extracted phase can effectively suppress OAM mode expansion caused by AT disturbance.