Multi-Level Feature Extraction Networks for Hyperspectral Image Classification
Hyperspectral image (HSI) classification plays a key role in the field of earth observation missions. Recently, transformer-based approaches have been widely used for HSI classification due to their ability to model long-range sequences. However, these methods face two main challenges. First, they t...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2024-02-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/16/3/590 |
| _version_ | 1797318229004124160 |
|---|---|
| author | Shaoyi Fang Xinyu Li Shimao Tian Weihao Chen Erlei Zhang |
| author_facet | Shaoyi Fang Xinyu Li Shimao Tian Weihao Chen Erlei Zhang |
| author_sort | Shaoyi Fang |
| collection | DOAJ |
| description | Hyperspectral image (HSI) classification plays a key role in the field of earth observation missions. Recently, transformer-based approaches have been widely used for HSI classification due to their ability to model long-range sequences. However, these methods face two main challenges. First, they treat HSI as linear vectors, disregarding their 3D attributes and spatial structure. Second, the repeated concatenation of encoders leads to information loss and gradient vanishing. To overcome these challenges, we propose a new solution called the multi-level feature extraction network (MLFEN). MLFEN consists of two sub-networks: the hybrid convolutional attention module (HCAM) and the enhanced dense vision transformer (EDVT). HCAM incorporates a band shift strategy to eliminate the edge effect of convolution and utilizes hybrid convolutional blocks to capture the 3D properties and spatial structure of HSI. Additionally, an attention module is introduced to identify strongly discriminative features. EDVT reconfigures the organization of original encoders by incorporating dense connections and adaptive feature fusion components, enabling faster propagation of information and mitigating the problem of gradient vanishing. Furthermore, we propose a novel sparse loss function to better fit the data distribution. Extensive experiments conducted on three public datasets demonstrate the significant advancements achieved by MLFEN. |
| first_indexed | 2024-03-08T03:49:23Z |
| format | Article |
| id | doaj.art-031ce165fd654575ac7c443ebac95916 |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-08T03:49:23Z |
| publishDate | 2024-02-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-031ce165fd654575ac7c443ebac959162024-02-09T15:21:35ZengMDPI AGRemote Sensing2072-42922024-02-0116359010.3390/rs16030590Multi-Level Feature Extraction Networks for Hyperspectral Image ClassificationShaoyi Fang0Xinyu Li1Shimao Tian2Weihao Chen3Erlei Zhang4School of Information Engineering, Northwest A&F University, Xi’an 712100, ChinaSchool of Information Engineering, Northwest A&F University, Xi’an 712100, ChinaSchool of Information Engineering, Northwest A&F University, Xi’an 712100, ChinaSchool of Information Engineering, Northwest A&F University, Xi’an 712100, ChinaSchool of Information Engineering, Northwest A&F University, Xi’an 712100, ChinaHyperspectral image (HSI) classification plays a key role in the field of earth observation missions. Recently, transformer-based approaches have been widely used for HSI classification due to their ability to model long-range sequences. However, these methods face two main challenges. First, they treat HSI as linear vectors, disregarding their 3D attributes and spatial structure. Second, the repeated concatenation of encoders leads to information loss and gradient vanishing. To overcome these challenges, we propose a new solution called the multi-level feature extraction network (MLFEN). MLFEN consists of two sub-networks: the hybrid convolutional attention module (HCAM) and the enhanced dense vision transformer (EDVT). HCAM incorporates a band shift strategy to eliminate the edge effect of convolution and utilizes hybrid convolutional blocks to capture the 3D properties and spatial structure of HSI. Additionally, an attention module is introduced to identify strongly discriminative features. EDVT reconfigures the organization of original encoders by incorporating dense connections and adaptive feature fusion components, enabling faster propagation of information and mitigating the problem of gradient vanishing. Furthermore, we propose a novel sparse loss function to better fit the data distribution. Extensive experiments conducted on three public datasets demonstrate the significant advancements achieved by MLFEN.https://www.mdpi.com/2072-4292/16/3/590hyperspectral image classificationconvolutional neural networksvision transformer |
| spellingShingle | Shaoyi Fang Xinyu Li Shimao Tian Weihao Chen Erlei Zhang Multi-Level Feature Extraction Networks for Hyperspectral Image Classification Remote Sensing hyperspectral image classification convolutional neural networks vision transformer |
| title | Multi-Level Feature Extraction Networks for Hyperspectral Image Classification |
| title_full | Multi-Level Feature Extraction Networks for Hyperspectral Image Classification |
| title_fullStr | Multi-Level Feature Extraction Networks for Hyperspectral Image Classification |
| title_full_unstemmed | Multi-Level Feature Extraction Networks for Hyperspectral Image Classification |
| title_short | Multi-Level Feature Extraction Networks for Hyperspectral Image Classification |
| title_sort | multi level feature extraction networks for hyperspectral image classification |
| topic | hyperspectral image classification convolutional neural networks vision transformer |
| url | https://www.mdpi.com/2072-4292/16/3/590 |
| work_keys_str_mv | AT shaoyifang multilevelfeatureextractionnetworksforhyperspectralimageclassification AT xinyuli multilevelfeatureextractionnetworksforhyperspectralimageclassification AT shimaotian multilevelfeatureextractionnetworksforhyperspectralimageclassification AT weihaochen multilevelfeatureextractionnetworksforhyperspectralimageclassification AT erleizhang multilevelfeatureextractionnetworksforhyperspectralimageclassification |