Interference Type Recognition in Spaceborne SARs Image Based on Deep CNN Model

Interference has an adverse impact on spaceborne SAR image interpretation, and mainly originates from terrestrial radio emitters. In recent years, a new type of mutual terrain scattered interference (MTSI) originating from other satellites also draw great attention, whose signal characteristics and resulting image artifacts are totally different from terrestrial interference. Determining the presence and type of interference in SAR images is an indispensable step before interference mitigation. In this paper, a novel interference recognition model using a deep convolutional neural network combined with attention mechanism is proposed. Combined with the attention mechanism, Resnet focuses on the changes in local interference regions and extracts the interference features in the image to distinguish interference. The distinction of MTSI and different kinds of terrestrial interference can be used as pre-processing before interference suppression. The experiments show that the proposed method outperforms other models on real measured Sentinel-1A data.