H-RNet: hybrid relation network for few-shot learning-based hyperspectral image classification.

Liu, Xiaoyong; Dong, Ziyang; Li, Huihui; Ren, Jinchang; Zhao, Huimin; Li, Hao; Chen, Weiqi; Xiao, Zhanhao

Abstract

Deep network models rely on sufficient training samples to perform reasonably well, which has inevitably constrained their application in classification of hyperspectral images (HSIs) due to the limited availability of labeled data. To tackle this particular challenge, we propose a hybrid relation network, H-RNet, by combining three-dimensional (3-D) convolution neural networks (CNN) and two-dimensional (2-D) CNN to extract the spectral–spatial features whilst reducing the complexity of the network. In an end-to-end relation learning module, the sample pairing approach can effectively alleviate the problem of few labeled samples and learn correlations between samples more accurately for more effective classification. Experimental results on three publicly available datasets have fully demonstrated the superior performance of the proposed model in comparison to a few state-of-the-art methods.

Citation

LIU, X., DONG, Z., LI, H., REN, J., ZHAO, H., LI, H., CHEN, W. and XIAO, Z. 2023. H-RNet: hybrid relation network for few-shot learning-based hyperspectral image classification. Remote sensing [online], 15(10), article 2497. Available from: https://doi.org/10.3390/rs15102497

• Journal Article Type: Article
• Acceptance Date: May 5, 2023
• Online Publication Date: May 9, 2023
• Publication Date: May 31, 2023
• Deposit Date: May 25, 2023
• Publicly Available Date: May 25, 2023
• Journal: Remote sensing
• Electronic ISSN: 2072-4292
• Publisher: MDPI
• Peer Reviewed: Peer Reviewed
• Volume: 15
• Issue: 10
• Article Number: 2497
• DOI: https://doi.org/10.3390/rs15102497
• Keywords: HSI classification; Few-shot learning; Relation network; Transfer learning
• Public URL: https://rgu-repository.worktribe.com/output/1961812