Water index and Swin Transformer Ensemble (WISTE) for water body extraction from multispectral remote sensing images
Automatic surface water body mapping using remote sensing technology is greatly meaningful for studying inland water dynamics at regional to global scales. Convolutional neural networks (CNN) have become an efficient semantic segmentation technique for the interpretation of remote sensing images. Ho...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Taylor & Francis Group
2023-12-01
|
| Series: | GIScience & Remote Sensing |
| Subjects: | |
| Online Access: | http://dx.doi.org/10.1080/15481603.2023.2251704 |
| _version_ | 1797679027333365760 |
|---|---|
| author | Donghui Ma Liguang Jiang Jie Li Yun Shi |
| author_facet | Donghui Ma Liguang Jiang Jie Li Yun Shi |
| author_sort | Donghui Ma |
| collection | DOAJ |
| description | Automatic surface water body mapping using remote sensing technology is greatly meaningful for studying inland water dynamics at regional to global scales. Convolutional neural networks (CNN) have become an efficient semantic segmentation technique for the interpretation of remote sensing images. However, the receptive field value of a CNN is restricted by the convolutional kernel size because the network only focuses on local features. The Swin Transformer has recently demonstrated its outstanding performance in computer vision tasks, and it could be useful for processing multispectral remote sensing images. In this article, a Water Index and Swin Transformer Ensemble (WISTE) method for automatic water body extraction is proposed. First, a dual-branch encoder architecture is designed for the Swin Transformer, aggregating the global semantic information and pixel neighbor relationships captured by fully convolutional networks (FCN) and multihead self-attention. Second, to prevent the Swin Transformer from ignoring multispectral information, we construct a prediction map ensemble module. The predictions of the Swin Transformer and the Normalized Difference Water Index (NDWI) are combined by a Bayesian averaging strategy. Finally, the experimental results obtained on two distinct datasets demonstrate that the WISTE has advantages over other segmentation methods and achieves the best results. The method proposed in this study can be used for improving regional to continental surface water mapping and related hydrological studies. |
| first_indexed | 2024-03-11T23:08:26Z |
| format | Article |
| id | doaj.art-6f8741a8805344288a1a8027a68a20a6 |
| institution | Directory Open Access Journal |
| issn | 1548-1603 1943-7226 |
| language | English |
| last_indexed | 2024-03-11T23:08:26Z |
| publishDate | 2023-12-01 |
| publisher | Taylor & Francis Group |
| record_format | Article |
| series | GIScience & Remote Sensing |
| spelling | doaj.art-6f8741a8805344288a1a8027a68a20a62023-09-21T12:43:10ZengTaylor & Francis GroupGIScience & Remote Sensing1548-16031943-72262023-12-0160110.1080/15481603.2023.22517042251704Water index and Swin Transformer Ensemble (WISTE) for water body extraction from multispectral remote sensing imagesDonghui Ma0Liguang Jiang1Jie Li2Yun Shi3Geovis Spatial Technology Co.LtdSchool of Environmental Science and Engineering, Southern University of Science and TechnologyXi’an Surveying and Mapping InstituteXi’an University of Science and TechnologyAutomatic surface water body mapping using remote sensing technology is greatly meaningful for studying inland water dynamics at regional to global scales. Convolutional neural networks (CNN) have become an efficient semantic segmentation technique for the interpretation of remote sensing images. However, the receptive field value of a CNN is restricted by the convolutional kernel size because the network only focuses on local features. The Swin Transformer has recently demonstrated its outstanding performance in computer vision tasks, and it could be useful for processing multispectral remote sensing images. In this article, a Water Index and Swin Transformer Ensemble (WISTE) method for automatic water body extraction is proposed. First, a dual-branch encoder architecture is designed for the Swin Transformer, aggregating the global semantic information and pixel neighbor relationships captured by fully convolutional networks (FCN) and multihead self-attention. Second, to prevent the Swin Transformer from ignoring multispectral information, we construct a prediction map ensemble module. The predictions of the Swin Transformer and the Normalized Difference Water Index (NDWI) are combined by a Bayesian averaging strategy. Finally, the experimental results obtained on two distinct datasets demonstrate that the WISTE has advantages over other segmentation methods and achieves the best results. The method proposed in this study can be used for improving regional to continental surface water mapping and related hydrological studies.http://dx.doi.org/10.1080/15481603.2023.2251704deep learningwater body extractionswin transformerwater indexdual-encoder |
| spellingShingle | Donghui Ma Liguang Jiang Jie Li Yun Shi Water index and Swin Transformer Ensemble (WISTE) for water body extraction from multispectral remote sensing images GIScience & Remote Sensing deep learning water body extraction swin transformer water index dual-encoder |
| title | Water index and Swin Transformer Ensemble (WISTE) for water body extraction from multispectral remote sensing images |
| title_full | Water index and Swin Transformer Ensemble (WISTE) for water body extraction from multispectral remote sensing images |
| title_fullStr | Water index and Swin Transformer Ensemble (WISTE) for water body extraction from multispectral remote sensing images |
| title_full_unstemmed | Water index and Swin Transformer Ensemble (WISTE) for water body extraction from multispectral remote sensing images |
| title_short | Water index and Swin Transformer Ensemble (WISTE) for water body extraction from multispectral remote sensing images |
| title_sort | water index and swin transformer ensemble wiste for water body extraction from multispectral remote sensing images |
| topic | deep learning water body extraction swin transformer water index dual-encoder |
| url | http://dx.doi.org/10.1080/15481603.2023.2251704 |
| work_keys_str_mv | AT donghuima waterindexandswintransformerensemblewisteforwaterbodyextractionfrommultispectralremotesensingimages AT liguangjiang waterindexandswintransformerensemblewisteforwaterbodyextractionfrommultispectralremotesensingimages AT jieli waterindexandswintransformerensemblewisteforwaterbodyextractionfrommultispectralremotesensingimages AT yunshi waterindexandswintransformerensemblewisteforwaterbodyextractionfrommultispectralremotesensingimages |