ISRNet: An Effective Network for SAR Interference Suppression and Recognition

Abstract

Synthetic aperture radar (SAR) is an all-weather, high-resolution sensor, but the presence of many kinds of interference in space can affect its mapping and survey capabilities. To address this problem, we propose an innovative network (ISRNet) for SAR interference suppression and recognition. Firstly, the SAR image model is proposed, including the target image, interference image, and background noise image. Secondly, ISRNet is composed of two parts, the interference suppression (IS) part and interference recognition (IR part), where IS part is based on Encoder-Decoder Block (EDB) and Feature Extraction Module (FEM) as the main framework to suppress interference. After that, we propose a novel interference recognition strategy: IR part to directly recognize the interference image to increase the recognition accuracy. Four types of interference data are adopted to verify the capability of ISRNet, and the results show that ISRNet has good suppression performance and high accuracy of interference recognition.