基于深度学习网络的单目标SAR图像生成研究  

Single-targetSAR Image Generation Via Deep Learning Networks

作　　者：王雪松 游俊[1,2] 强勇 WANG Xuesong;YOU Jun;QIANG Yong(Xi'an Electronic Engineering Research Institute,Xi'an 710100;Xi'an Key Laboratory of RF Integrated Simulation,Xi'an 710100)

机构地区：[1]西安电子工程研究所,西安710100 [2]西安射频综合仿真重点实验室,西安710100

出　　处：《火控雷达技术》2025年第1期16-24,共9页Fire Control Radar Technology

摘　　要：传统的合成孔径雷达(Synthetic Aperture Radar,SAR)图像生成方法往往存在着建模复杂,计算量大,花费时间长等问题。同时在面对非合作目标时,无法获得目标的精确模型以及全方位角的SAR图像,这对后续的研究形成了很大阻碍。本文对原始CWGAN-GP网络的损失函数进行了优化,并提出了一种改进的特征编码方法用于描述SAR图像中目标的方位角特征。实验生成了间隔为1°的360幅仿真SAR图像,扩充了SAR图像数据集,最后将扩充后的图像数据用于识别网络的训练,通过识别正确率的提高验证了利用深度学习网络生成仿真SAR图像数据的可行性。Traditional methods for generating Synthetic Aperture Radar(SAR)images often face challenges such as complex modeling,large computational demands,and long processing times.When dealing with non-cooperative targets,it is impossible to obtain accurate models and omni-directional SAR images of the target,which greatly hinders subsequent research.This article optimizes the loss function of the original CWGAN-GP network and proposes a feature encoding method to describe the azimuth features of targets in SAR images.Experiments generate 360 simulated SAR images with an interval of 1°,expanding the SAR image dataset.Finally,the expanded image dataset is used for training the recognition network.The improvement in recognition accuracy validates the feasibility of generating simulated SAR image data.

关 键 词：合成孔径雷达 生成对抗网络 数据增强 图像生成 

分 类 号：TN957.52[电子电信—信号与信息处理]