Random Shuffling Data for Hyperspectral Image Classification with Siamese and Knowledge Distillation Network
Hyperspectral images (HSIs) are characterized by hundreds of spectral bands. The goal of HSI is to associate the pixel with a corresponding category label by analyzing subtle differences in the spectrum. Due to their excellent local context modeling capabilities, Convolutional Neural Network (CNN)-b...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-08-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/15/16/4078 |
| _version_ | 1797583254716416000 |
|---|---|
| author | Zhen Yang Ying Cao Xin Zhou Junya Liu Tao Zhang Jinsheng Ji |
| author_facet | Zhen Yang Ying Cao Xin Zhou Junya Liu Tao Zhang Jinsheng Ji |
| author_sort | Zhen Yang |
| collection | DOAJ |
| description | Hyperspectral images (HSIs) are characterized by hundreds of spectral bands. The goal of HSI is to associate the pixel with a corresponding category label by analyzing subtle differences in the spectrum. Due to their excellent local context modeling capabilities, Convolutional Neural Network (CNN)-based methods are often adopted to complete the classification task. To verify whether the patch-data-based CNN methods depend on the homogeneity of patch data during the training process in HSI classification, we designed a random shuffling strategy to disrupt the data homogeneity of the patch data, which is randomly assigning the pixels from the original dataset to other positions to form a new dataset. Based on this random shuffling strategy, we propose a sub-branch to extract features on the reconstructed dataset and fuse the loss rates (RFL). The loss rate calculated by RFL in the new patch data is cross combined with the loss value calculated by another sub-branch in the original patch data. Moreover, we construct a new hyperspectral classification network based on the Siamese and Knowledge Distillation Network (SKDN) that can improve the classification accuracy on randomly shuffled data. In addition, RFL is introduced into the original model for hyperspectral classification tasks in the original dataset. The experimental results show that the improved model is also better than the original model, which indicates that RFL is effective and feasible. Experiments on four real-world datasets show that, as the proportion of randomly shuffling data increases, the latest patch-data-based CNN methods cannot extract more abundant local contextual information for HSI classification, while the proposed sub-branch RFL can alleviate this problem and improve the network’s recognition ability. |
| first_indexed | 2024-03-10T23:36:16Z |
| format | Article |
| id | doaj.art-106dcbff970e4ff89be7b43bd47a9e46 |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-10T23:36:16Z |
| publishDate | 2023-08-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-106dcbff970e4ff89be7b43bd47a9e462023-11-19T02:54:14ZengMDPI AGRemote Sensing2072-42922023-08-011516407810.3390/rs15164078Random Shuffling Data for Hyperspectral Image Classification with Siamese and Knowledge Distillation NetworkZhen Yang0Ying Cao1Xin Zhou2Junya Liu3Tao Zhang4Jinsheng Ji5The School of Information and Electromechanical Engineering, Jiangxi Science and Technology Normal University, Nanchang 330013, ChinaThe School of Information and Electromechanical Engineering, Jiangxi Science and Technology Normal University, Nanchang 330013, ChinaThe School of Information and Electromechanical Engineering, Jiangxi Science and Technology Normal University, Nanchang 330013, ChinaThe School of Information and Electromechanical Engineering, Jiangxi Science and Technology Normal University, Nanchang 330013, ChinaShanghai Key Laboratory of Intelligent Sensing and Recognition, Shanghai Jiao Tong University, Shanghai 200240, ChinaSchool of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798, SingaporeHyperspectral images (HSIs) are characterized by hundreds of spectral bands. The goal of HSI is to associate the pixel with a corresponding category label by analyzing subtle differences in the spectrum. Due to their excellent local context modeling capabilities, Convolutional Neural Network (CNN)-based methods are often adopted to complete the classification task. To verify whether the patch-data-based CNN methods depend on the homogeneity of patch data during the training process in HSI classification, we designed a random shuffling strategy to disrupt the data homogeneity of the patch data, which is randomly assigning the pixels from the original dataset to other positions to form a new dataset. Based on this random shuffling strategy, we propose a sub-branch to extract features on the reconstructed dataset and fuse the loss rates (RFL). The loss rate calculated by RFL in the new patch data is cross combined with the loss value calculated by another sub-branch in the original patch data. Moreover, we construct a new hyperspectral classification network based on the Siamese and Knowledge Distillation Network (SKDN) that can improve the classification accuracy on randomly shuffled data. In addition, RFL is introduced into the original model for hyperspectral classification tasks in the original dataset. The experimental results show that the improved model is also better than the original model, which indicates that RFL is effective and feasible. Experiments on four real-world datasets show that, as the proportion of randomly shuffling data increases, the latest patch-data-based CNN methods cannot extract more abundant local contextual information for HSI classification, while the proposed sub-branch RFL can alleviate this problem and improve the network’s recognition ability.https://www.mdpi.com/2072-4292/15/16/4078patch-data-based CNN methodlocal contextual informationrandom shuffling strategynetwork SKDN consists of two sub-branchesHSI classification |
| spellingShingle | Zhen Yang Ying Cao Xin Zhou Junya Liu Tao Zhang Jinsheng Ji Random Shuffling Data for Hyperspectral Image Classification with Siamese and Knowledge Distillation Network Remote Sensing patch-data-based CNN method local contextual information random shuffling strategy network SKDN consists of two sub-branches HSI classification |
| title | Random Shuffling Data for Hyperspectral Image Classification with Siamese and Knowledge Distillation Network |
| title_full | Random Shuffling Data for Hyperspectral Image Classification with Siamese and Knowledge Distillation Network |
| title_fullStr | Random Shuffling Data for Hyperspectral Image Classification with Siamese and Knowledge Distillation Network |
| title_full_unstemmed | Random Shuffling Data for Hyperspectral Image Classification with Siamese and Knowledge Distillation Network |
| title_short | Random Shuffling Data for Hyperspectral Image Classification with Siamese and Knowledge Distillation Network |
| title_sort | random shuffling data for hyperspectral image classification with siamese and knowledge distillation network |
| topic | patch-data-based CNN method local contextual information random shuffling strategy network SKDN consists of two sub-branches HSI classification |
| url | https://www.mdpi.com/2072-4292/15/16/4078 |
| work_keys_str_mv | AT zhenyang randomshufflingdataforhyperspectralimageclassificationwithsiameseandknowledgedistillationnetwork AT yingcao randomshufflingdataforhyperspectralimageclassificationwithsiameseandknowledgedistillationnetwork AT xinzhou randomshufflingdataforhyperspectralimageclassificationwithsiameseandknowledgedistillationnetwork AT junyaliu randomshufflingdataforhyperspectralimageclassificationwithsiameseandknowledgedistillationnetwork AT taozhang randomshufflingdataforhyperspectralimageclassificationwithsiameseandknowledgedistillationnetwork AT jinshengji randomshufflingdataforhyperspectralimageclassificationwithsiameseandknowledgedistillationnetwork |