基于生成对抗网络的遥感影像场景分类

Scene Classification of Remote Sensing Images Based on Generative Adversarial Networks

针对遥感图像标签样本数据不足的问题,本文通过改进生成对抗网络分类方法,以更好地利用标签信息增强目标域类间的区分度,并利用AID、NWPU-RESISC45、UCMerced_LandUse和WHU-RS19数据集构建源域数据集和目标域数据集,并在构建的数据集上进行了实验。结果表明本文方法对目标域有较好的分类效果。

Aiming to insufficient labeled images in remote sensing,a new method was proposed to make better use of label information and enhance the differentiation between target domain classes by improving classification methods based on generative adversarial networks.Then,the method was experimented on the source domain dataset and target domain dataset which were built by selecting some categories from AID,NWPU-RESISC45,UCMerced LandUse and WHU-RS19 datasets,and results show that it has performance better for the target domain dataset.