Auxiliary Classifier GAN-Based Super-resolution Microwave Coincidence Imaging

Microwave coincidence imaging (MCI) has attracted more and more attention, but the number of the low correlated measurement modes is limited leading to a limited imaging performance. The unsupervised data augmentation can be achieved by the Generative Adversarial Networks (GAN) to address the issue of limited samples. In this paper, an Auxiliary Classifier GAN-based super-resolution imaging method for MCI is proposed. To achieve the classification of ten uniformly distributed target points within the 3dB beamwidth (i.e. 10 times super-resolution imaging), a training dataset is designed based on the spatial correlation characteristics of the reference radiation field in MCI systems. Experimental results demonstrate that the proposed approach can effectively generate high-quality samples to meet classification requirements. Compared with other methods, the accuracy and classification efficiency can be improved.