LG mode optical angular momentum state identification method and system based on deep learning

Abstract

The invention discloses an LG mode optical angular momentum state recognition method and system based on deep learning. The method comprises the steps of LG mode extraction, wherein an LG mode is generated from an input optical image or signal through an SLM; data preprocessing and discretization processing: preprocessing LG mode data generated through SLM, and discretizing continuous data to input a deep learning model; constructing a dual-stage ResNet50 deep learning model; inputting the discretized LG mode data into the deep learning model for training and identification; in the first stage, the model classifies the basic layer to define categories of OAM data sets in different dimensions; in the second stage, detailed sub-classification is carried out based on the initial classification; and model verification and optimization: verifying and optimizing the trained deep learning model. According to the method, the customized dual-stage neural network architecture is adopted, and discretization processing is performed on the transverse spatial mode state space of the structured light, so that excellent OAM state recognition accuracy is realized.

Images