Representation and Classification of Auroral Images Based on Convolutional Neural Networks
Auroral forms are correlated with certain physical processes in the magnetosphere and ionosphere. It is, therefore, desirable to automatically classify the vast amount of observed auroral images and make large statistical studies. The key problem in classification tasks is image representation. In t...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2020-01-01
|
| Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/8970288/ |
| _version_ | 1819150149438406656 |
|---|---|
| author | Qiuju Yang Penghui Zhou |
| author_facet | Qiuju Yang Penghui Zhou |
| author_sort | Qiuju Yang |
| collection | DOAJ |
| description | Auroral forms are correlated with certain physical processes in the magnetosphere and ionosphere. It is, therefore, desirable to automatically classify the vast amount of observed auroral images and make large statistical studies. The key problem in classification tasks is image representation. In this article, using the adaptive feature learning ability of convolutional neural networks, an end-to-end auroral image classification network is proposed, which automatically classifies the auroral images observed at the Chinese Yellow River Station into four classes: arc, drapery corona, radial corona, and hotspot corona. Based on the AlexNet, our method exploits the advanced spatial transformer network (STN) and large margin Softmax (L-Softmax) loss function to extract auroral features. STN is able to learn invariance to translation, scaling, and rotation, whereas L-Softmax increases the difficulty of auroral feature learning so that it encourages the intraclass compactness and interclass separability between learned features. The proposed method was validated on the auroral image datasets by supervised classification, image retrieval, and statistical analysis of the temporal occurrence distributions of the four auroral categories. Experimental results showed that after trained on the winter auroral observations in 2003, the proposed model achieves an average classification accuracy of 93.7% on the auroral data of the following five winters (2004-2009) while maintaining high efficiency, which is superior to the previously reported articles. |
| first_indexed | 2024-12-22T14:12:55Z |
| format | Article |
| id | doaj.art-bb38b94fec5c4786a6b9113cf60cac1a |
| institution | Directory Open Access Journal |
| issn | 2151-1535 |
| language | English |
| last_indexed | 2024-12-22T14:12:55Z |
| publishDate | 2020-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| spelling | doaj.art-bb38b94fec5c4786a6b9113cf60cac1a2022-12-21T18:23:11ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing2151-15352020-01-011352353410.1109/JSTARS.2020.29692458970288Representation and Classification of Auroral Images Based on Convolutional Neural NetworksQiuju Yang0https://orcid.org/0000-0001-9773-3806Penghui Zhou1School of Physics and Information Technology, Shaanxi Normal University, Xi'an, ChinaSchool of Physics and Information Technology, Shaanxi Normal University, Xi'an, ChinaAuroral forms are correlated with certain physical processes in the magnetosphere and ionosphere. It is, therefore, desirable to automatically classify the vast amount of observed auroral images and make large statistical studies. The key problem in classification tasks is image representation. In this article, using the adaptive feature learning ability of convolutional neural networks, an end-to-end auroral image classification network is proposed, which automatically classifies the auroral images observed at the Chinese Yellow River Station into four classes: arc, drapery corona, radial corona, and hotspot corona. Based on the AlexNet, our method exploits the advanced spatial transformer network (STN) and large margin Softmax (L-Softmax) loss function to extract auroral features. STN is able to learn invariance to translation, scaling, and rotation, whereas L-Softmax increases the difficulty of auroral feature learning so that it encourages the intraclass compactness and interclass separability between learned features. The proposed method was validated on the auroral image datasets by supervised classification, image retrieval, and statistical analysis of the temporal occurrence distributions of the four auroral categories. Experimental results showed that after trained on the winter auroral observations in 2003, the proposed model achieves an average classification accuracy of 93.7% on the auroral data of the following five winters (2004-2009) while maintaining high efficiency, which is superior to the previously reported articles.https://ieeexplore.ieee.org/document/8970288/Auroral image classificationconvolutional neural network (CNN)large margin Softmax (L-Softmax) loss functionspatial transformer network (STN) |
| spellingShingle | Qiuju Yang Penghui Zhou Representation and Classification of Auroral Images Based on Convolutional Neural Networks IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Auroral image classification convolutional neural network (CNN) large margin Softmax (L-Softmax) loss function spatial transformer network (STN) |
| title | Representation and Classification of Auroral Images Based on Convolutional Neural Networks |
| title_full | Representation and Classification of Auroral Images Based on Convolutional Neural Networks |
| title_fullStr | Representation and Classification of Auroral Images Based on Convolutional Neural Networks |
| title_full_unstemmed | Representation and Classification of Auroral Images Based on Convolutional Neural Networks |
| title_short | Representation and Classification of Auroral Images Based on Convolutional Neural Networks |
| title_sort | representation and classification of auroral images based on convolutional neural networks |
| topic | Auroral image classification convolutional neural network (CNN) large margin Softmax (L-Softmax) loss function spatial transformer network (STN) |
| url | https://ieeexplore.ieee.org/document/8970288/ |
| work_keys_str_mv | AT qiujuyang representationandclassificationofauroralimagesbasedonconvolutionalneuralnetworks AT penghuizhou representationandclassificationofauroralimagesbasedonconvolutionalneuralnetworks |