无线电工程2025，Vol.55Issue(3)：580-587,8.DOI:10.3969/j.issn.1003-3106.2025.03.015
基于CDCGAN的SAR图像数据增广
SAR Image Augmentation Method Based on CDCGAN
赵竹新 1范纯卓 1刘艳博 1冯彦卿 1王海强1
作者信息
- 1. 北京市遥感信息研究所,北京 100011
- 折叠
摘要
Abstract
To address the problems of high cost and lack of diversity of Synthetic Aperture Radar(SAR)image acquisition which affect the effect of image interpretation,on the basis of existing Deep Convolutional Generative Adversarial Network(DCGAN),this paper proposed a model based on conditional DCGAN(DCGAN with conditional input)which realizes SAR image augmentation with customized azimuth/pitch/oblique angle and greatly expands the existing simulation dataset.Also,an evaluation index system of SAR augmented images is established to evaluate the quality of SAR augmented images objectively.The results show that the proposed method can successfully achieve the expansion of SAR image samples with multiple customized angles with satisfied quality and lower time cost,which is of positive significance for enhancing the diversity of SAR image angles.关键词
合成孔径雷达仿真/语义可控条件生成对抗网络/卷积/角度信息可控Key words
SAR simulation/CDCGAN/convolution/angle information customized分类
信息技术与安全科学引用本文复制引用
赵竹新,范纯卓,刘艳博,冯彦卿,王海强..基于CDCGAN的SAR图像数据增广[J].无线电工程,2025,55(3):580-587,8.
