A Multipath and Multiscale Siamese Network Based on Spatial-Spectral Features for Few-Shot Hyperspectral Image Classification
Deep learning has been demonstrated to be a powerful nonlinear modeling method with end-to-end optimization capabilities for hyperspectral Images (HSIs). However, in real classification cases, obtaining labeled samples is often time-consuming and labor-intensive, resulting in few-shot training sampl...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-09-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/15/18/4391 |
| _version_ | 1827723867850801152 |
|---|---|
| author | Jinghui Yang Jia Qin Jinxi Qian Anqi Li Liguo Wang |
| author_facet | Jinghui Yang Jia Qin Jinxi Qian Anqi Li Liguo Wang |
| author_sort | Jinghui Yang |
| collection | DOAJ |
| description | Deep learning has been demonstrated to be a powerful nonlinear modeling method with end-to-end optimization capabilities for hyperspectral Images (HSIs). However, in real classification cases, obtaining labeled samples is often time-consuming and labor-intensive, resulting in few-shot training samples. Based on this issue, a multipath and multiscale Siamese network based on spatial-spectral features for few-shot hyperspectral image classification (MMSN) is proposed. To conduct classification with few-shot training samples, a Siamese network framework with low dependence on sample information is adopted. In one subnetwork, a spatial attention module (DCAM), which combines dilated convolution and cosine similarity to comprehensively consider spatial-spectral weights, is designed first. Then, we propose a residual-dense hybrid module (RDHM), which merges three-path features, including grouped convolution-based local residual features, global residual features and global dense features. The RDHM can effectively propagate and utilize different layers of features and enhance the expression ability of these features. Finally, we construct a multikernel depth feature extraction module (MDFE) that performs multiscale convolutions with multikernel and hierarchical skip connections on the feature scales to improve the ability of the network to capture details. Extensive experimental evidence shows that the proposed MMSN method exhibits a superior performance on few-shot training samples, and its classification results are better than those of other state-of-the-art classification methods. |
| first_indexed | 2024-03-10T22:06:49Z |
| format | Article |
| id | doaj.art-e158d7effb554239b12c130ede064dc2 |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-10T22:06:49Z |
| publishDate | 2023-09-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-e158d7effb554239b12c130ede064dc22023-11-19T12:47:02ZengMDPI AGRemote Sensing2072-42922023-09-011518439110.3390/rs15184391A Multipath and Multiscale Siamese Network Based on Spatial-Spectral Features for Few-Shot Hyperspectral Image ClassificationJinghui Yang0Jia Qin1Jinxi Qian2Anqi Li3Liguo Wang4School of Information Engineering, China University of Geosciences (Beijing), Beijing 100083, ChinaSchool of Information Engineering, China University of Geosciences (Beijing), Beijing 100083, ChinaInstitute of Telecommunication and Navigation Satellites, China Academy of Space Technology, Beijing 100094, ChinaSchool of Information Engineering, China University of Geosciences (Beijing), Beijing 100083, ChinaCollege of Information and Communication Engineering, Dalian Minzu University, Dalian 116600, ChinaDeep learning has been demonstrated to be a powerful nonlinear modeling method with end-to-end optimization capabilities for hyperspectral Images (HSIs). However, in real classification cases, obtaining labeled samples is often time-consuming and labor-intensive, resulting in few-shot training samples. Based on this issue, a multipath and multiscale Siamese network based on spatial-spectral features for few-shot hyperspectral image classification (MMSN) is proposed. To conduct classification with few-shot training samples, a Siamese network framework with low dependence on sample information is adopted. In one subnetwork, a spatial attention module (DCAM), which combines dilated convolution and cosine similarity to comprehensively consider spatial-spectral weights, is designed first. Then, we propose a residual-dense hybrid module (RDHM), which merges three-path features, including grouped convolution-based local residual features, global residual features and global dense features. The RDHM can effectively propagate and utilize different layers of features and enhance the expression ability of these features. Finally, we construct a multikernel depth feature extraction module (MDFE) that performs multiscale convolutions with multikernel and hierarchical skip connections on the feature scales to improve the ability of the network to capture details. Extensive experimental evidence shows that the proposed MMSN method exhibits a superior performance on few-shot training samples, and its classification results are better than those of other state-of-the-art classification methods.https://www.mdpi.com/2072-4292/15/18/4391hyperspectral image (HSI)Siamese networkfew-shot training samplesclassificationmultiscale |
| spellingShingle | Jinghui Yang Jia Qin Jinxi Qian Anqi Li Liguo Wang A Multipath and Multiscale Siamese Network Based on Spatial-Spectral Features for Few-Shot Hyperspectral Image Classification Remote Sensing hyperspectral image (HSI) Siamese network few-shot training samples classification multiscale |
| title | A Multipath and Multiscale Siamese Network Based on Spatial-Spectral Features for Few-Shot Hyperspectral Image Classification |
| title_full | A Multipath and Multiscale Siamese Network Based on Spatial-Spectral Features for Few-Shot Hyperspectral Image Classification |
| title_fullStr | A Multipath and Multiscale Siamese Network Based on Spatial-Spectral Features for Few-Shot Hyperspectral Image Classification |
| title_full_unstemmed | A Multipath and Multiscale Siamese Network Based on Spatial-Spectral Features for Few-Shot Hyperspectral Image Classification |
| title_short | A Multipath and Multiscale Siamese Network Based on Spatial-Spectral Features for Few-Shot Hyperspectral Image Classification |
| title_sort | multipath and multiscale siamese network based on spatial spectral features for few shot hyperspectral image classification |
| topic | hyperspectral image (HSI) Siamese network few-shot training samples classification multiscale |
| url | https://www.mdpi.com/2072-4292/15/18/4391 |
| work_keys_str_mv | AT jinghuiyang amultipathandmultiscalesiamesenetworkbasedonspatialspectralfeaturesforfewshothyperspectralimageclassification AT jiaqin amultipathandmultiscalesiamesenetworkbasedonspatialspectralfeaturesforfewshothyperspectralimageclassification AT jinxiqian amultipathandmultiscalesiamesenetworkbasedonspatialspectralfeaturesforfewshothyperspectralimageclassification AT anqili amultipathandmultiscalesiamesenetworkbasedonspatialspectralfeaturesforfewshothyperspectralimageclassification AT liguowang amultipathandmultiscalesiamesenetworkbasedonspatialspectralfeaturesforfewshothyperspectralimageclassification AT jinghuiyang multipathandmultiscalesiamesenetworkbasedonspatialspectralfeaturesforfewshothyperspectralimageclassification AT jiaqin multipathandmultiscalesiamesenetworkbasedonspatialspectralfeaturesforfewshothyperspectralimageclassification AT jinxiqian multipathandmultiscalesiamesenetworkbasedonspatialspectralfeaturesforfewshothyperspectralimageclassification AT anqili multipathandmultiscalesiamesenetworkbasedonspatialspectralfeaturesforfewshothyperspectralimageclassification AT liguowang multipathandmultiscalesiamesenetworkbasedonspatialspectralfeaturesforfewshothyperspectralimageclassification |