Feature recognition of a 2D array vortex interferogram using a convolutional neural network

Abstract

A vortex array has important applications in scenarios where multiple vortex elements with the same or different topological charges are required simultaneously. Therefore, the detection of the vortex array is vital. Here, the interferogram between the off-axis Walsh-phase plate and the vortex array is first obtained and then decoded through a convolution neural network (CNN), which can simultaneously determine the topological charge, chirality, and the initial angle. Both the theory and experiment prove that a CNN has a remarkable effect on the classification and detection of vortex arrays.