Research on Single SAR Deception Jamming Template Augmentation Scheme Based on Generative Adversarial Networks
-
摘要: 对单张合成孔径雷达(SAR)欺骗干扰模板进行样本增广，生成高质量的SAR欺骗干扰模版库，有助于进行快速有效的SAR欺骗干扰。目前SAR欺骗干扰模板的样本增广方案由于缺乏相干斑噪声，导致生成模板的真实性较低，同时生成的模板图像与原图相似性较低。针对该问题，本文提出了一种基于生成对抗网络的样本增广方案，在网络中考虑了相干斑噪声的影响，并使用注意力机制模块、残差密集模块、多尺度模块来提高网络对特征的提取能力。在MSTAR数据集上的实验表明，本方案生成的图像具有与原始图像更加相似的图像特征，并且含有相似的相干斑噪声特征，具有更高的真实性，由此验证了方法的有效性。Abstract: Augmenting samples of single Synthetic Aperture Radar (SAR) deception jamming templates to generate a high-quality SAR deception jamming template library facilitates rapid and effective SAR deception jamming. Currently, the sample augmentation scheme for SAR deception jamming templates lacks coherent speckle noise, resulting in lower authenticity of generated templates, and simultaneously, lower similarity between generated template images and original images. To address this issue, this paper proposes a sample augmentation scheme based on Generative Adversarial Networks (GANs), considering the influence of coherent speckle noise within the network, and utilizing attention mechanism modules, residual dense mod-ules, and multi-scale modules to enhance the network's feature extraction capability. Experiments on the MSTAR dataset demonstrate that the images generated by this scheme exhibit more similar image features to the original images and contain similar coherent speckle noise features, thereby confirming the effectiveness of the proposed method.
-
计量
- 文章访问数: 140
- HTML全文浏览量: 2
- PDF下载量: 21
下载:
