Few-Shot Unsupervised Domain Adaptation Based on Refined Bi-Directional Prototypical Contrastive Learning for Cross-Scene Hyperspectral Image Classification

Hyperspectral image cross-scene classification (HSICC) tasks are confronted with tremendous challenges due to spectral shift phenomena across scenes and the tough work of obtaining labels. Unsupervised domain adaptation has proven its effectiveness in tackling this issue, but it has a fatal limitati...

Full description

Saved in:
Bibliographic Details
Main Authors: Xuebin Tang, Hanyi Shi, Chunchao Li, Cheng Jiang, Xiaoxiong Zhang, Lingbin Zeng, Xiaolei Zhou
Format: Article
Language:English
Published: MDPI AG 2025-07-01
Series:Remote Sensing
Subjects:
Online Access:https://www.mdpi.com/2072-4292/17/13/2305
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849427575499128832
author Xuebin Tang
Hanyi Shi
Chunchao Li
Cheng Jiang
Xiaoxiong Zhang
Lingbin Zeng
Xiaolei Zhou
author_facet Xuebin Tang
Hanyi Shi
Chunchao Li
Cheng Jiang
Xiaoxiong Zhang
Lingbin Zeng
Xiaolei Zhou
author_sort Xuebin Tang
collection DOAJ
description Hyperspectral image cross-scene classification (HSICC) tasks are confronted with tremendous challenges due to spectral shift phenomena across scenes and the tough work of obtaining labels. Unsupervised domain adaptation has proven its effectiveness in tackling this issue, but it has a fatal limitation of intending to narrow the disparity between source and target domains by utilizing fully labeled source data and unlabeled target data. However, it is costly even to attain labels from source domains in many cases, rendering sufficient labeling as used in prior work impractical. In this work, we investigate an extreme and realistic scenario where unsupervised domain adaptation methods encounter sparsely labeled source data when handling HSICC tasks, namely, few-shot unsupervised domain adaptation. We propose an end-to-end refined bi-directional prototypical contrastive learning (RBPCL) framework for overcoming the HSICC problem with only a few labeled samples in the source domain. RBPCL captures category-level semantic features of hyperspectral data and performs feature alignment through in-domain refined prototypical self-supervised learning and bi-directional cross-domain prototypical contrastive learning, respectively. Furthermore, our framework introduces the class-balanced multicentric dynamic prototype strategy to generate more robust and representative prototypes. To facilitate prototype contrastive learning, we employ a Siamese-style distance metric loss function to aggregate intra-class features while increasing the discrepancy of inter-class features. Finally, extensive experiments and ablation analysis implemented on two public cross-scene data pairs and three pairs of self-collected ultralow-altitude hyperspectral datasets under different illumination conditions verify the effectiveness of our method, which will further enhance the practicality of hyperspectral intelligent sensing technology.
format Article
id doaj-art-297a70106cf04807b6049a9cbc297471
institution Kabale University
issn 2072-4292
language English
publishDate 2025-07-01
publisher MDPI AG
record_format Article
series Remote Sensing
spelling doaj-art-297a70106cf04807b6049a9cbc2974712025-08-20T03:28:59ZengMDPI AGRemote Sensing2072-42922025-07-011713230510.3390/rs17132305Few-Shot Unsupervised Domain Adaptation Based on Refined Bi-Directional Prototypical Contrastive Learning for Cross-Scene Hyperspectral Image ClassificationXuebin Tang0Hanyi Shi1Chunchao Li2Cheng Jiang3Xiaoxiong Zhang4Lingbin Zeng5Xiaolei Zhou6National University of Defense Technology, Changsha 410073, ChinaNational University of Defense Technology, Changsha 410073, ChinaNational University of Defense Technology, Changsha 410073, ChinaNational University of Defense Technology, Changsha 410073, ChinaNational University of Defense Technology, Changsha 410073, ChinaNational University of Defense Technology, Changsha 410073, ChinaNational University of Defense Technology, Changsha 410073, ChinaHyperspectral image cross-scene classification (HSICC) tasks are confronted with tremendous challenges due to spectral shift phenomena across scenes and the tough work of obtaining labels. Unsupervised domain adaptation has proven its effectiveness in tackling this issue, but it has a fatal limitation of intending to narrow the disparity between source and target domains by utilizing fully labeled source data and unlabeled target data. However, it is costly even to attain labels from source domains in many cases, rendering sufficient labeling as used in prior work impractical. In this work, we investigate an extreme and realistic scenario where unsupervised domain adaptation methods encounter sparsely labeled source data when handling HSICC tasks, namely, few-shot unsupervised domain adaptation. We propose an end-to-end refined bi-directional prototypical contrastive learning (RBPCL) framework for overcoming the HSICC problem with only a few labeled samples in the source domain. RBPCL captures category-level semantic features of hyperspectral data and performs feature alignment through in-domain refined prototypical self-supervised learning and bi-directional cross-domain prototypical contrastive learning, respectively. Furthermore, our framework introduces the class-balanced multicentric dynamic prototype strategy to generate more robust and representative prototypes. To facilitate prototype contrastive learning, we employ a Siamese-style distance metric loss function to aggregate intra-class features while increasing the discrepancy of inter-class features. Finally, extensive experiments and ablation analysis implemented on two public cross-scene data pairs and three pairs of self-collected ultralow-altitude hyperspectral datasets under different illumination conditions verify the effectiveness of our method, which will further enhance the practicality of hyperspectral intelligent sensing technology.https://www.mdpi.com/2072-4292/17/13/2305remote sensinghyperspectral imaging cross-scene classification (HSICC)few-shot unsupervised domain adaptation (FUDA)refined bi-directional prototypical contrastive learning (RBPCL)self-collected ultralow-altitude hyperspectral datasets
spellingShingle Xuebin Tang
Hanyi Shi
Chunchao Li
Cheng Jiang
Xiaoxiong Zhang
Lingbin Zeng
Xiaolei Zhou
Few-Shot Unsupervised Domain Adaptation Based on Refined Bi-Directional Prototypical Contrastive Learning for Cross-Scene Hyperspectral Image Classification
Remote Sensing
remote sensing
hyperspectral imaging cross-scene classification (HSICC)
few-shot unsupervised domain adaptation (FUDA)
refined bi-directional prototypical contrastive learning (RBPCL)
self-collected ultralow-altitude hyperspectral datasets
title Few-Shot Unsupervised Domain Adaptation Based on Refined Bi-Directional Prototypical Contrastive Learning for Cross-Scene Hyperspectral Image Classification
title_full Few-Shot Unsupervised Domain Adaptation Based on Refined Bi-Directional Prototypical Contrastive Learning for Cross-Scene Hyperspectral Image Classification
title_fullStr Few-Shot Unsupervised Domain Adaptation Based on Refined Bi-Directional Prototypical Contrastive Learning for Cross-Scene Hyperspectral Image Classification
title_full_unstemmed Few-Shot Unsupervised Domain Adaptation Based on Refined Bi-Directional Prototypical Contrastive Learning for Cross-Scene Hyperspectral Image Classification
title_short Few-Shot Unsupervised Domain Adaptation Based on Refined Bi-Directional Prototypical Contrastive Learning for Cross-Scene Hyperspectral Image Classification
title_sort few shot unsupervised domain adaptation based on refined bi directional prototypical contrastive learning for cross scene hyperspectral image classification
topic remote sensing
hyperspectral imaging cross-scene classification (HSICC)
few-shot unsupervised domain adaptation (FUDA)
refined bi-directional prototypical contrastive learning (RBPCL)
self-collected ultralow-altitude hyperspectral datasets
url https://www.mdpi.com/2072-4292/17/13/2305
work_keys_str_mv AT xuebintang fewshotunsuperviseddomainadaptationbasedonrefinedbidirectionalprototypicalcontrastivelearningforcrossscenehyperspectralimageclassification
AT hanyishi fewshotunsuperviseddomainadaptationbasedonrefinedbidirectionalprototypicalcontrastivelearningforcrossscenehyperspectralimageclassification
AT chunchaoli fewshotunsuperviseddomainadaptationbasedonrefinedbidirectionalprototypicalcontrastivelearningforcrossscenehyperspectralimageclassification
AT chengjiang fewshotunsuperviseddomainadaptationbasedonrefinedbidirectionalprototypicalcontrastivelearningforcrossscenehyperspectralimageclassification
AT xiaoxiongzhang fewshotunsuperviseddomainadaptationbasedonrefinedbidirectionalprototypicalcontrastivelearningforcrossscenehyperspectralimageclassification
AT lingbinzeng fewshotunsuperviseddomainadaptationbasedonrefinedbidirectionalprototypicalcontrastivelearningforcrossscenehyperspectralimageclassification
AT xiaoleizhou fewshotunsuperviseddomainadaptationbasedonrefinedbidirectionalprototypicalcontrastivelearningforcrossscenehyperspectralimageclassification