Deep Metric Learning with Online Hard Mining for Hyperspectral Classification
Recently, deep learning has developed rapidly, while it has also been quite successfully applied in the field of hyperspectral classification. Generally, training the parameters of a deep neural network to the best is the core step of a deep learning-based method, which usually requires a large numb...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-04-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/13/7/1368 |
| _version_ | 1797538989753761792 |
|---|---|
| author | Yanni Dong Cong Yang Yuxiang Zhang |
| author_facet | Yanni Dong Cong Yang Yuxiang Zhang |
| author_sort | Yanni Dong |
| collection | DOAJ |
| description | Recently, deep learning has developed rapidly, while it has also been quite successfully applied in the field of hyperspectral classification. Generally, training the parameters of a deep neural network to the best is the core step of a deep learning-based method, which usually requires a large number of labeled samples. However, in remote sensing analysis tasks, we only have limited labeled data because of the high cost of their collection. Therefore, in this paper, we propose a deep metric learning with online hard mining (DMLOHM) method for hyperspectral classification, which can maximize the inter-class distance and minimize the intra-class distance, utilizing a convolutional neural network (CNN) as an embedded network. First of all, we utilized the triplet network to learn better representations of raw data so that raw data were capable of having their dimensionality reduced. Afterward, an online hard mining method was used to mine the most valuable information from the limited hyperspectral data. To verify the performance of the proposed DMLOHM, we utilized three well-known hyperspectral datasets: Salinas Scene, Pavia University, and HyRANK for verification. Compared with CNN and DMLTN, the experimental results showed that the proposed method improved the classification accuracy from 0.13% to 4.03% with 85 labeled samples per class. |
| first_indexed | 2024-03-10T12:39:01Z |
| format | Article |
| id | doaj.art-cd100be56e8b432998efc652bed8b75e |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-10T12:39:01Z |
| publishDate | 2021-04-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-cd100be56e8b432998efc652bed8b75e2023-11-21T14:01:44ZengMDPI AGRemote Sensing2072-42922021-04-01137136810.3390/rs13071368Deep Metric Learning with Online Hard Mining for Hyperspectral ClassificationYanni Dong0Cong Yang1Yuxiang Zhang2Hubei Subsurface Multi-Scale Imaging Key Laboratory, Institute of Geophysics and Geomatics, China University of Geosciences, Wuhan 430074, ChinaHubei Subsurface Multi-Scale Imaging Key Laboratory, Institute of Geophysics and Geomatics, China University of Geosciences, Wuhan 430074, ChinaHubei Subsurface Multi-Scale Imaging Key Laboratory, Institute of Geophysics and Geomatics, China University of Geosciences, Wuhan 430074, ChinaRecently, deep learning has developed rapidly, while it has also been quite successfully applied in the field of hyperspectral classification. Generally, training the parameters of a deep neural network to the best is the core step of a deep learning-based method, which usually requires a large number of labeled samples. However, in remote sensing analysis tasks, we only have limited labeled data because of the high cost of their collection. Therefore, in this paper, we propose a deep metric learning with online hard mining (DMLOHM) method for hyperspectral classification, which can maximize the inter-class distance and minimize the intra-class distance, utilizing a convolutional neural network (CNN) as an embedded network. First of all, we utilized the triplet network to learn better representations of raw data so that raw data were capable of having their dimensionality reduced. Afterward, an online hard mining method was used to mine the most valuable information from the limited hyperspectral data. To verify the performance of the proposed DMLOHM, we utilized three well-known hyperspectral datasets: Salinas Scene, Pavia University, and HyRANK for verification. Compared with CNN and DMLTN, the experimental results showed that the proposed method improved the classification accuracy from 0.13% to 4.03% with 85 labeled samples per class.https://www.mdpi.com/2072-4292/13/7/1368hyperspectral classificationdeep metric learningonline hard mining |
| spellingShingle | Yanni Dong Cong Yang Yuxiang Zhang Deep Metric Learning with Online Hard Mining for Hyperspectral Classification Remote Sensing hyperspectral classification deep metric learning online hard mining |
| title | Deep Metric Learning with Online Hard Mining for Hyperspectral Classification |
| title_full | Deep Metric Learning with Online Hard Mining for Hyperspectral Classification |
| title_fullStr | Deep Metric Learning with Online Hard Mining for Hyperspectral Classification |
| title_full_unstemmed | Deep Metric Learning with Online Hard Mining for Hyperspectral Classification |
| title_short | Deep Metric Learning with Online Hard Mining for Hyperspectral Classification |
| title_sort | deep metric learning with online hard mining for hyperspectral classification |
| topic | hyperspectral classification deep metric learning online hard mining |
| url | https://www.mdpi.com/2072-4292/13/7/1368 |
| work_keys_str_mv | AT yannidong deepmetriclearningwithonlinehardminingforhyperspectralclassification AT congyang deepmetriclearningwithonlinehardminingforhyperspectralclassification AT yuxiangzhang deepmetriclearningwithonlinehardminingforhyperspectralclassification |