Deep transformer and few‐shot learning for hyperspectral image classification
Abstract Recently, deep learning has achieved considerable results in the hyperspectral image (HSI) classification. However, most available deep networks require ample and authentic samples to better train the models, which is expensive and inefficient in practical tasks. Existing few‐shot learning...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2023-12-01
|
| Series: | CAAI Transactions on Intelligence Technology |
| Subjects: | |
| Online Access: | https://doi.org/10.1049/cit2.12181 |
| _version_ | 1827577063065780224 |
|---|---|
| author | Qiong Ran Yonghao Zhou Danfeng Hong Meiqiao Bi Li Ni Xuan Li Muhammad Ahmad |
| author_facet | Qiong Ran Yonghao Zhou Danfeng Hong Meiqiao Bi Li Ni Xuan Li Muhammad Ahmad |
| author_sort | Qiong Ran |
| collection | DOAJ |
| description | Abstract Recently, deep learning has achieved considerable results in the hyperspectral image (HSI) classification. However, most available deep networks require ample and authentic samples to better train the models, which is expensive and inefficient in practical tasks. Existing few‐shot learning (FSL) methods generally ignore the potential relationships between non‐local spatial samples that would better represent the underlying features of HSI. To solve the above issues, a novel deep transformer and few‐shot learning (DT‐FSL) classification framework is proposed, attempting to realize fine‐grained classification of HSI with only a few‐shot instances. Specifically, the spatial attention and spectral query modules are introduced to overcome the constraint of the convolution kernel and consider the information between long‐distance location (non‐local) samples to reduce the uncertainty of classes. Next, the network is trained with episodes and task‐based learning strategies to learn a metric space, which can continuously enhance its modelling capability. Furthermore, the developed approach combines the advantages of domain adaptation to reduce the variation in inter‐domain distribution and realize distribution alignment. On three publicly available HSI data, extensive experiments have indicated that the proposed DT‐FSL yields better results concerning state‐of‐the‐art algorithms. |
| first_indexed | 2024-03-08T21:21:24Z |
| format | Article |
| id | doaj.art-88ac2b1d3e0f43ada688bec8ca1989e3 |
| institution | Directory Open Access Journal |
| issn | 2468-2322 |
| language | English |
| last_indexed | 2024-03-08T21:21:24Z |
| publishDate | 2023-12-01 |
| publisher | Wiley |
| record_format | Article |
| series | CAAI Transactions on Intelligence Technology |
| spelling | doaj.art-88ac2b1d3e0f43ada688bec8ca1989e32023-12-21T09:45:29ZengWileyCAAI Transactions on Intelligence Technology2468-23222023-12-01841323133610.1049/cit2.12181Deep transformer and few‐shot learning for hyperspectral image classificationQiong Ran0Yonghao Zhou1Danfeng Hong2Meiqiao Bi3Li Ni4Xuan Li5Muhammad Ahmad6College of Information Science and Technology Beijing University of Chemical Technology Beijing ChinaCollege of Information Science and Technology Beijing University of Chemical Technology Beijing ChinaKey Laboratory of Computational Optical Imaging Technology Aerospace Information Research Institute Chinese Academy of Sciences Beijing ChinaKey Laboratory of Computational Optical Imaging Technology Aerospace Information Research Institute Chinese Academy of Sciences Beijing ChinaKey Laboratory of Computational Optical Imaging Technology Aerospace Information Research Institute Chinese Academy of Sciences Beijing ChinaChina Greatwall Technology Group Co., Ltd Shenzhen ChinaDepartment of Computer Science National University of Computer and Emerging Sciences Chiniot PakistanAbstract Recently, deep learning has achieved considerable results in the hyperspectral image (HSI) classification. However, most available deep networks require ample and authentic samples to better train the models, which is expensive and inefficient in practical tasks. Existing few‐shot learning (FSL) methods generally ignore the potential relationships between non‐local spatial samples that would better represent the underlying features of HSI. To solve the above issues, a novel deep transformer and few‐shot learning (DT‐FSL) classification framework is proposed, attempting to realize fine‐grained classification of HSI with only a few‐shot instances. Specifically, the spatial attention and spectral query modules are introduced to overcome the constraint of the convolution kernel and consider the information between long‐distance location (non‐local) samples to reduce the uncertainty of classes. Next, the network is trained with episodes and task‐based learning strategies to learn a metric space, which can continuously enhance its modelling capability. Furthermore, the developed approach combines the advantages of domain adaptation to reduce the variation in inter‐domain distribution and realize distribution alignment. On three publicly available HSI data, extensive experiments have indicated that the proposed DT‐FSL yields better results concerning state‐of‐the‐art algorithms.https://doi.org/10.1049/cit2.12181deep learningfeature extractionhyperspectralimage classification |
| spellingShingle | Qiong Ran Yonghao Zhou Danfeng Hong Meiqiao Bi Li Ni Xuan Li Muhammad Ahmad Deep transformer and few‐shot learning for hyperspectral image classification CAAI Transactions on Intelligence Technology deep learning feature extraction hyperspectral image classification |
| title | Deep transformer and few‐shot learning for hyperspectral image classification |
| title_full | Deep transformer and few‐shot learning for hyperspectral image classification |
| title_fullStr | Deep transformer and few‐shot learning for hyperspectral image classification |
| title_full_unstemmed | Deep transformer and few‐shot learning for hyperspectral image classification |
| title_short | Deep transformer and few‐shot learning for hyperspectral image classification |
| title_sort | deep transformer and few shot learning for hyperspectral image classification |
| topic | deep learning feature extraction hyperspectral image classification |
| url | https://doi.org/10.1049/cit2.12181 |
| work_keys_str_mv | AT qiongran deeptransformerandfewshotlearningforhyperspectralimageclassification AT yonghaozhou deeptransformerandfewshotlearningforhyperspectralimageclassification AT danfenghong deeptransformerandfewshotlearningforhyperspectralimageclassification AT meiqiaobi deeptransformerandfewshotlearningforhyperspectralimageclassification AT lini deeptransformerandfewshotlearningforhyperspectralimageclassification AT xuanli deeptransformerandfewshotlearningforhyperspectralimageclassification AT muhammadahmad deeptransformerandfewshotlearningforhyperspectralimageclassification |