基于卷积注意力的逆合成孔径雷达成像方法  

Inverse Synthetic Aperture Radar imaging method based on Convolutional Attention

作　　者：李成坤 刘霖 刘亚波 张跃 LI Chengkun;LIU Lin;LIU Yabo;ZHANG Yue(Aerospace Information Research Institute,Chinese Academy of Sciences,Beijing 100094,China;School of Electronic,Electrical and Communication Engineering,University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院空天信息创新研究院,北京100094 [2]中国科学院大学电子电气与通信工程学院,北京100049

出　　处：《电子设计工程》2024年第14期59-63,共5页Electronic Design Engineering

摘　　要：近年来,卷积神经网络由于其强大的特征提取能力和对数据的灵活表达,已成功应用于逆合成孔径雷达(ISAR)的稀疏成像。然而常规的卷积神经网络对于特征给予相同的权重,缺乏聚焦重要信息和抑制冗余信息的能力,导致成像效果欠佳。为了解决这个问题,提出了基于卷积注意力的逆合成孔径雷达成像方法,在网络中加入卷积注意力模块,在通道和空间两个维度上进行特征细化,提高网络对于特征的表达能力,并且可以在残差连接中减少原始特征中的伪影干扰,提高成像质量。实验结果表明,基于卷积注意力的成像方法可以重建高质量ISAR图像,且相较于全卷积神经网络成像方法具有更好的成像效果。所成图像熵值从0.355 9降到0.221 2,证明了所提方法的有效性。In recent years,convolutional neural network has been successfully applied to sparse imaging of Inverse Synthetic Aperture Radar(ISAR)due to its powerful feature extraction ability and flexible representation of data.However,conventional Convolutional Neural Networks give the same weight to features,lacks the ability to focus on important information and suppress redundant information,resulting in poor imaging results.In order to solve this problem,an inverse synthetic aperture radar imaging method based on the convolutional attention is proposed.The convolutional block attention module is added to the network to refine the features in the channel and space dimensions,improve the ability of the network to express the features,and reduce the artifacts interference in the original features in the residual connection to improve the imaging quality.The results of experiment show that the imaging method based on Convolutional Attention can reconstruct high-quality ISAR images and has better imaging effect than the Full Convolutional Neural Network imaging method.The image entropy is reduced from 0.3559 to 0.2212,which proves the effectiveness of the proposed method.

关 键 词：逆合成孔径雷达成像 深度学习 卷积神经网络 卷积注意力 

分 类 号：TN957.52[电子电信—信号与信息处理]