Super-Resolution Microwave Coincidence Imaging Based on Squeeze-and-Excitation CNN
Published in 2026 IEEE 14th Asia-Pacific Conference on Antennas and Propagation (APCAP), 2026
In this paper, a super-resolution microwave coincidence imaging (MCI) method based on a Squeeze-and-Excitatoin Convolutional Neural Network (SE-CNN) is proposed. The proposed SE-CNN employs a modified Gram-Schmidt orthogonalization algorithm to extract amplitude and phase features of the radiation field and is trained on a two-channel dataset incorporating multi-target interference and ambient noise. It dynamically weights feature channels to predict the primary target’s position, with the complex scattering coefficient inverted via least-squares. Numerical simulations demonstrate 10-fold super-resolution in distinguishing targets within the 3 dB beamwidth at a 20 dB signal-to-noise ratio, achieving high accuracy and robustness under strong interference.
Citation: P. Li, M. Zhao, D. Li, Q. Yang, P. Qi, and S. Zhu, "Super-resolution microwave coincidence imaging based on squeeze-and-excitation CNN," 2026 IEEE 14th Asia-Pacific Conference on Antennas and Propagation (APCAP), Chengdu, China, 2026, pp. 875-876.
