A Robust Dynamic Classifier Selection Approach for Hyperspectral Images with Imprecise Label Information
Supervised hyperspectral image (HSI) classification relies on accurate label information. However, it is not always possible to collect perfectly accurate labels for training samples. This motivates the development of classifiers that are sufficiently robust to some reasonable amounts of errors in d...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2020-09-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/20/18/5262 |
| _version_ | 1797553657562005504 |
|---|---|
| author | Meizhu Li Shaoguang Huang Jasper De Bock Gert de Cooman Aleksandra Pižurica |
| author_facet | Meizhu Li Shaoguang Huang Jasper De Bock Gert de Cooman Aleksandra Pižurica |
| author_sort | Meizhu Li |
| collection | DOAJ |
| description | Supervised hyperspectral image (HSI) classification relies on accurate label information. However, it is not always possible to collect perfectly accurate labels for training samples. This motivates the development of classifiers that are sufficiently robust to some reasonable amounts of errors in data labels. Despite the growing importance of this aspect, it has not been sufficiently studied in the literature yet. In this paper, we analyze the effect of erroneous sample labels on probability distributions of the principal components of HSIs, and provide in this way a statistical analysis of the resulting uncertainty in classifiers. Building on the theory of imprecise probabilities, we develop a novel robust dynamic classifier selection (R-DCS) model for data classification with erroneous labels. Particularly, spectral and spatial features are extracted from HSIs to construct two individual classifiers for the dynamic selection, respectively. The proposed R-DCS model is based on the robustness of the classifiers’ predictions: the extent to which a classifier can be altered without changing its prediction. We provide three possible selection strategies for the proposed model with different computational complexities and apply them on three benchmark data sets. Experimental results demonstrate that the proposed model outperforms the individual classifiers it selects from and is more robust to errors in labels compared to widely adopted approaches. |
| first_indexed | 2024-03-10T16:19:42Z |
| format | Article |
| id | doaj.art-8c22ac86abd8417ca6d6c2b99bd38741 |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-10T16:19:42Z |
| publishDate | 2020-09-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-8c22ac86abd8417ca6d6c2b99bd387412023-11-20T13:46:25ZengMDPI AGSensors1424-82202020-09-012018526210.3390/s20185262A Robust Dynamic Classifier Selection Approach for Hyperspectral Images with Imprecise Label InformationMeizhu Li0Shaoguang Huang1Jasper De Bock2Gert de Cooman3Aleksandra Pižurica4GAIM, Department of Telecommunications and Information Processing, Ghent University, 9000 Gent, BelgiumGAIM, Department of Telecommunications and Information Processing, Ghent University, 9000 Gent, BelgiumFLip, Department of Electronics and Information Systems, Ghent University, 9052 Gent, BelgiumFLip, Department of Electronics and Information Systems, Ghent University, 9052 Gent, BelgiumGAIM, Department of Telecommunications and Information Processing, Ghent University, 9000 Gent, BelgiumSupervised hyperspectral image (HSI) classification relies on accurate label information. However, it is not always possible to collect perfectly accurate labels for training samples. This motivates the development of classifiers that are sufficiently robust to some reasonable amounts of errors in data labels. Despite the growing importance of this aspect, it has not been sufficiently studied in the literature yet. In this paper, we analyze the effect of erroneous sample labels on probability distributions of the principal components of HSIs, and provide in this way a statistical analysis of the resulting uncertainty in classifiers. Building on the theory of imprecise probabilities, we develop a novel robust dynamic classifier selection (R-DCS) model for data classification with erroneous labels. Particularly, spectral and spatial features are extracted from HSIs to construct two individual classifiers for the dynamic selection, respectively. The proposed R-DCS model is based on the robustness of the classifiers’ predictions: the extent to which a classifier can be altered without changing its prediction. We provide three possible selection strategies for the proposed model with different computational complexities and apply them on three benchmark data sets. Experimental results demonstrate that the proposed model outperforms the individual classifiers it selects from and is more robust to errors in labels compared to widely adopted approaches.https://www.mdpi.com/1424-8220/20/18/5262robust classificationdynamic classifier selectionhyperspectral imagesnoisy labelsimprecise probabilities |
| spellingShingle | Meizhu Li Shaoguang Huang Jasper De Bock Gert de Cooman Aleksandra Pižurica A Robust Dynamic Classifier Selection Approach for Hyperspectral Images with Imprecise Label Information Sensors robust classification dynamic classifier selection hyperspectral images noisy labels imprecise probabilities |
| title | A Robust Dynamic Classifier Selection Approach for Hyperspectral Images with Imprecise Label Information |
| title_full | A Robust Dynamic Classifier Selection Approach for Hyperspectral Images with Imprecise Label Information |
| title_fullStr | A Robust Dynamic Classifier Selection Approach for Hyperspectral Images with Imprecise Label Information |
| title_full_unstemmed | A Robust Dynamic Classifier Selection Approach for Hyperspectral Images with Imprecise Label Information |
| title_short | A Robust Dynamic Classifier Selection Approach for Hyperspectral Images with Imprecise Label Information |
| title_sort | robust dynamic classifier selection approach for hyperspectral images with imprecise label information |
| topic | robust classification dynamic classifier selection hyperspectral images noisy labels imprecise probabilities |
| url | https://www.mdpi.com/1424-8220/20/18/5262 |
| work_keys_str_mv | AT meizhuli arobustdynamicclassifierselectionapproachforhyperspectralimageswithimpreciselabelinformation AT shaoguanghuang arobustdynamicclassifierselectionapproachforhyperspectralimageswithimpreciselabelinformation AT jasperdebock arobustdynamicclassifierselectionapproachforhyperspectralimageswithimpreciselabelinformation AT gertdecooman arobustdynamicclassifierselectionapproachforhyperspectralimageswithimpreciselabelinformation AT aleksandrapizurica arobustdynamicclassifierselectionapproachforhyperspectralimageswithimpreciselabelinformation AT meizhuli robustdynamicclassifierselectionapproachforhyperspectralimageswithimpreciselabelinformation AT shaoguanghuang robustdynamicclassifierselectionapproachforhyperspectralimageswithimpreciselabelinformation AT jasperdebock robustdynamicclassifierselectionapproachforhyperspectralimageswithimpreciselabelinformation AT gertdecooman robustdynamicclassifierselectionapproachforhyperspectralimageswithimpreciselabelinformation AT aleksandrapizurica robustdynamicclassifierselectionapproachforhyperspectralimageswithimpreciselabelinformation |