Invariant semantic domain generalization shuffle network for cross-scene hyperspectral image classification

IF 7.5 1区 计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE
Jingpeng Gao, Xiangyu Ji, Fang Ye, Geng Chen
{"title":"Invariant semantic domain generalization shuffle network for cross-scene hyperspectral image classification","authors":"Jingpeng Gao,&nbsp;Xiangyu Ji,&nbsp;Fang Ye,&nbsp;Geng Chen","doi":"10.1016/j.eswa.2025.126818","DOIUrl":null,"url":null,"abstract":"<div><div>Cross-scene hyperspectral image classification is currently receiving widespread attention. However, domain adaptation-based methods usually perform domain alignment by accessing specific target scenes during training and require retraining for new scenes. In contrast, domain generalization only trains using the source domain and then gradually generalizes to unseen domains. However, existing methods based on domain generalization ignore the impact of domain invariant semantics on the invariant representation of the domain. To solve the above problem, an invariant semantic domain generalization shuffle network for cross-scene hyperspectral image classification is proposed, which follows a framework on the generative adversarial network. Feature style covariance in style and content randomization generator with invariant semantic features is designed to safely extend the style and content of features without changing the domain invariant semantics. We proposed a spatial shuffling discriminator, which can reduce the impact of special spatial relationships within the domain on class semantics. In addition, we proposed a dual sampling direct adversarial contrastive learning strategy. It uses a dual sampling in two-stage training design to prevent the model from lazily entering the local nash equilibrium point. And based on dual sampling, directly adversarial contrastive learning using clearer contrastive samples is used to reduce the difficulty of network training. We conduct extensive experiments on four datasets and demonstrate that the proposed method outperforms other current domain generalization methods. The code will be open source at <span><span>https://github.com/jixiangyu0501/ISDGS</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"273 ","pages":"Article 126818"},"PeriodicalIF":7.5000,"publicationDate":"2025-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Expert Systems with Applications","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0957417425004403","RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
引用次数: 0

Abstract

Cross-scene hyperspectral image classification is currently receiving widespread attention. However, domain adaptation-based methods usually perform domain alignment by accessing specific target scenes during training and require retraining for new scenes. In contrast, domain generalization only trains using the source domain and then gradually generalizes to unseen domains. However, existing methods based on domain generalization ignore the impact of domain invariant semantics on the invariant representation of the domain. To solve the above problem, an invariant semantic domain generalization shuffle network for cross-scene hyperspectral image classification is proposed, which follows a framework on the generative adversarial network. Feature style covariance in style and content randomization generator with invariant semantic features is designed to safely extend the style and content of features without changing the domain invariant semantics. We proposed a spatial shuffling discriminator, which can reduce the impact of special spatial relationships within the domain on class semantics. In addition, we proposed a dual sampling direct adversarial contrastive learning strategy. It uses a dual sampling in two-stage training design to prevent the model from lazily entering the local nash equilibrium point. And based on dual sampling, directly adversarial contrastive learning using clearer contrastive samples is used to reduce the difficulty of network training. We conduct extensive experiments on four datasets and demonstrate that the proposed method outperforms other current domain generalization methods. The code will be open source at https://github.com/jixiangyu0501/ISDGS.
求助全文
约1分钟内获得全文 求助全文
来源期刊
Expert Systems with Applications
Expert Systems with Applications 工程技术-工程:电子与电气
CiteScore
13.80
自引率
10.60%
发文量
2045
审稿时长
8.7 months
期刊介绍: Expert Systems With Applications is an international journal dedicated to the exchange of information on expert and intelligent systems used globally in industry, government, and universities. The journal emphasizes original papers covering the design, development, testing, implementation, and management of these systems, offering practical guidelines. It spans various sectors such as finance, engineering, marketing, law, project management, information management, medicine, and more. The journal also welcomes papers on multi-agent systems, knowledge management, neural networks, knowledge discovery, data mining, and other related areas, excluding applications to military/defense systems.
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信