Grating Lobe Suppression of Multichannel Forward-Looking SAR Based on Self-Supervised Contrastive Learning

Multichannel radar can resolve left/right ambiguity and has the potential of high-resolution forward-looking imaging. However, when only a limited number of channels are available, in order to have good performance of resolving left/right ambiguity, there is always grating lobe in the imaging results. In this paper, a scheme of grating lobe suppression based on self-supervised contrastive learning is proposed. Firstly, a generative adversarial network is employed to achieve the mapping from the source domain X (i.e. the imaging results with fewer channels) to the target domain Y (i.e. the imaging results with more channels). Then, a contrastive loss is introduced to ensure across-domain preservation of the content information like the distribution of ground objects, preventing the generator from making unnecessary changes in results when suppressing the grating lobe. At last, simulation results are given to illustrate the effectiveness of the proposed method.