Near‐Field Microwave Holographic Imaging Using a Metamaterial‐Based Diffraction Neural Network With Angular Spectrum‐Assisted Computation
Ying Li, Li Deng, Meijun Qu · Microwave and Optical Technology Letters · 2025
ABSTRACT This Letter focuses on the imaging capability of a diffraction neural network. By combining the Rayleigh–Sommerfeld diffraction equation with neural network connectivity rules, we design a diffraction neural network model for holographic imaging on a zero‐information input plane. The model is implemented using the angular spectrum method and fast Fourier transform algorithm. Training is performed via error backpropagation and a stochastic gradient descent algorithm. We analyze the theoretical influence of key network parameters and conduct simulation‐based studies to explore the effects of unit size, working wavelength, and diffraction propagation distance on imaging performance. Metamaterials are used to simulate Huygens diffraction neurons. A physical‐layer model of the diffraction neural network is constructed using metamaterial parameter mapping, and its effectiveness is evaluated through full‐wave electromagnetic simulations. The simulation results demonstrate the feasibility and potential of the proposed imaging system based on the diffraction neural network.