福建电脑2025，Vol.41Issue(7)：17-21,5.DOI:10.16707/j.cnki.fjpc.2025.07.004
面向遥感云去除的循环生成对抗网络技术研究
Research on Cycle Generative Adversarial Network Technology for Cloud Removal in Remote Sensing Images
摘要
Abstract
To address the impact of cloud cover on remote sensing image information extraction and analysis,this paper proposes a cyclic generative adversarial network structure for remote sensing cloud removal.Utilizing the advantages of pixel to pixel conversion in recurrent generative adversarial networks and introducing channel space attention modules to solve the problem of detail loss;By recalibrating the feature image information to enhance the feature fusion effect,and using block processing methods to improve the discriminative effect of the discriminator on image details and overall structure.Adopting gradient penalty and Wasserstein loss to prevent model collapse.The experimental results show that the peak signal-to-noise ratio of our method is 26.684,the structural similarity is 0.892,and the mean square error is 0.027,which is superior to other methods and can effectively remove clouds from remote sensing images.关键词
遥感影像/遥感云去除/循环生成对抗网络Key words
Remote Sensing Imagery/Remote Sensing Cloud Removal/Recurrent Generative Adversarial Network分类
信息技术与安全科学引用本文复制引用
古彭,邱晓凤..面向遥感云去除的循环生成对抗网络技术研究[J].福建电脑,2025,41(7):17-21,5.基金项目
本文得到福建省中青年教师教育科研项目(科技类)基金(No.JAT231217)、三明市引导性科技项目基金(No.2024-G-031)、校级课题(YJKJ2307B)资助. （科技类）
