基于DCGAN的SAR虚假目标图像仿真  

Deceptive Target SAR Image Simulation Based on Deep Convolutional Generative Adversarial Network

作　　者：卢庆林 叶伟 李国靖 LU Qinglin;YE Wei;LI Guojing(Space Engineering University,Beijing 101416,China)

机构地区：[1]航天工程大学,北京101416

出　　处：《电子信息对抗技术》2020年第2期57-61,65,共6页Electronic Information Warfare Technology

摘　　要：利用预设的虚假目标图像模板进行信号反演,是SAR欺骗干扰的一种有效手段。为了提供逼真度高的虚假目标图像模板,提出一种基于深度卷积生成式对抗网络(DCGAN)的SAR虚假目标图像生成方法。通过分析模板图像的目视效果、统计特性和在目标识别网络中的表现,评估虚假目标模板图像的逼真度。实验证明,DCGAN能生成特征明显的目标图像,在目视效果和统计特性方面与真实目标图像近似度较高。目标识别网络的检测效果也体现出模板图像具有较高可信度,证明该仿真方法生成虚假目标图像模板的有效性。The signal inversion of deceptive target image templates is an effective means of SAR deceptive jamming. In order to effectively combat the SAR target recognition system based on the deep learning method, a SAR deceptive target image generation method based on Deep Convolutional Generative Adversarial Network(DCGAN) is proposed. The fidelity of the deceptive target image is evaluated by analyzing the visual effects, statistical characteristics, and performance of the template image in the target recognition network. The experiment result proves that DCGAN can generate SAR images with obvious features, and target images have high approximation to real target images in terms of visual effects and statistical characteristics. The result of recognition network expresses that the template image has high credibility, which proves that the simulation method of deceptive target images has high superiority.

关 键 词：合成孔径雷达 深度卷积生成式对抗网络 虚假目标图像模板 图像质量 可信度 

分 类 号：TN972.3[电子电信—信号与信息处理]