基于区域特征细化感知学习的星载SAR图像有源压制干扰抑制方法

Active Blanket Jamming Suppression Method for Spaceborne SAR Images Based on Regional Feature Refinement Perceptual Learning

星载合成孔径雷达(SAR)系统常受到强电磁干扰而导致成像质量下降,但现有基于图像域的干扰抑制方法易造成图像失真、纹理细节信息丢失等难题。针对上述问题,该文提出了一种基于区域特征细化感知学习的星载SAR图像有源压制干扰抑制方法。首先,建立了星载SAR图像域有源压制干扰信号和图像模型;其次,设计一种基于区域特征感知的高精度干扰识别网络,利用高效通道注意力机制,提取SAR图像有源压制干扰图样特征,可以有效识别SAR图像干扰区域;然后,构建一种基于SAR图像和压制干扰特征联合学习的多元区域特征细化干扰抑制网络,将SAR图像切分为多元区域,采用多模块协同处理多元区域上的压制干扰特征,实现复杂场景条件下SAR图像有源压制干扰的精细化抑制。最后,构建SAR图像有源压制干扰仿真数据集,且采用哨兵1号实测数据进行实验验证分析。实验结果表明所提方法能有效识别和抑制星载SAR图像多种典型有源压制干扰。

Spaceborne Synthetic Aperture Radar(SAR)systems are often subject to strong electromagnetic interference,resulting in imaging quality degradation.However,existing image domain-based interference suppression methods are prone to image distortion and loss of texture detail information,among other difficulties.To address these problems,this paper proposes a method for suppressing active suppression interferences inspaceborne SAR images based on perceptual learning of regional feature refinement.First,an active suppression interference signal and image model is established in the spaceborne SAR image domain.Second,a high-precision interference recognition network based on regional feature perception is designed to extract the active suppression interference pattern features of the involved SAR image using an efficient channel attention mechanism,consequently resulting in effective recognition of the interference region of the SAR image.Third,a multivariate regional feature refinement interference suppression network is constructed based on the joint learning of the SAR image and suppression interference features,which are combined to form the SAR image and suppression interference pattern.A feature refinement interference suppression network is then constructed based on the joint learning of the SAR image and suppression interference feature.The network slices the SAR image into multivariate regions,and adopts multi-module collaborative processing of suppression interference features on the multivariate regions to realize refined suppression of the active suppression interference of the SAR image under complex conditions.Finally,a simulation dataset of SAR image active suppression interference is constructed,and the evaluated Sentinel-1 data are used for experimental verification and analysis.The experimental results show that the proposed method can effectively recognize and suppress various typical active suppression interferences in spaceborne SAR images.