Efficient recognition of composite vortex beams with multi-mode in atmospheric turbulence based on convolutional neural network

In recent years, vortex beams have been widely used in optical communication because of their characteristics. However, it is noted that the information encoding of a single-mode vortex beam is limited by the size of physical device. To resolve the information limitation problem associated with single-mode vortex beams, an idea of multi-beam switching superposition can be applied to composite vortex beam with multi-mode. Here, a convolutional neural network is combined to simulate and recognize the specific modes of composite vortex beams. Moreover, taking into account the atmospheric turbulence environment, the recognition effect is numerically simulated under three different levels of turbulence intensity. Finally, a compensation scheme is proposed for strong turbulence, whose recognition rate is upto 97.82% in harsh environments. This finding attests to the efficiency of our proposed technique. These results are expected to generate novel concepts for enhancing optical communication capacity and quality in the future.