A Novel Ground-Based Cloud Image Segmentation Method Based on a Multibranch Asymmetric Convolution Module and Attention Mechanism
Cloud segmentation is a fundamental step in accurately acquiring cloud cover. However, due to the nonrigid structures of clouds, traditional cloud segmentation methods perform worse than expected. In this paper, a novel deep convolutional neural network (CNN) named MA-SegCloud is proposed for segmen...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-08-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/14/16/3970 |
| _version_ | 1827598480890134528 |
|---|---|
| author | Liwen Zhang Wenhao Wei Bo Qiu Ali Luo Mingru Zhang Xiaotong Li |
| author_facet | Liwen Zhang Wenhao Wei Bo Qiu Ali Luo Mingru Zhang Xiaotong Li |
| author_sort | Liwen Zhang |
| collection | DOAJ |
| description | Cloud segmentation is a fundamental step in accurately acquiring cloud cover. However, due to the nonrigid structures of clouds, traditional cloud segmentation methods perform worse than expected. In this paper, a novel deep convolutional neural network (CNN) named MA-SegCloud is proposed for segmenting cloud images based on a multibranch asymmetric convolution module (MACM) and an attention mechanism. The MACM is composed of asymmetric convolution, depth-separable convolution, and a squeeze-and-excitation module (SEM). The MACM not only enables the network to capture more contextual information in a larger area but can also adaptively adjust the feature channel weights. The attention mechanisms SEM and convolutional block attention module (CBAM) in the network can strengthen useful features for cloud image segmentation. As a result, MA-SegCloud achieves a 96.9% accuracy, 97.0% precision, 97.0% recall, 97.0% F-score, 3.1% error rate, and 94.0% mean intersection-over-union (MIoU) on the Singapore Whole-sky Nychthemeron Image Segmentation (SWINySEG) dataset. Extensive evaluations demonstrate that MA-SegCloud performs favorably against state-of-the-art cloud image segmentation methods. |
| first_indexed | 2024-03-09T03:53:33Z |
| format | Article |
| id | doaj.art-7dcb27bc03b943d18746572a0659c568 |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-09T03:53:33Z |
| publishDate | 2022-08-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-7dcb27bc03b943d18746572a0659c5682023-12-03T14:24:26ZengMDPI AGRemote Sensing2072-42922022-08-011416397010.3390/rs14163970A Novel Ground-Based Cloud Image Segmentation Method Based on a Multibranch Asymmetric Convolution Module and Attention MechanismLiwen Zhang0Wenhao Wei1Bo Qiu2Ali Luo3Mingru Zhang4Xiaotong Li5School of Electronic Information Engineering, Hebei University of Technology, Tianjin 300401, ChinaSchool of Electronic Information Engineering, Hebei University of Technology, Tianjin 300401, ChinaSchool of Electronic Information Engineering, Hebei University of Technology, Tianjin 300401, ChinaCAS Key Laboratory of Optical Astronomy, National Astronomical Observatories, Beijing 100101, ChinaSchool of Electronic Information Engineering, Hebei University of Technology, Tianjin 300401, ChinaSchool of Electronic Information Engineering, Hebei University of Technology, Tianjin 300401, ChinaCloud segmentation is a fundamental step in accurately acquiring cloud cover. However, due to the nonrigid structures of clouds, traditional cloud segmentation methods perform worse than expected. In this paper, a novel deep convolutional neural network (CNN) named MA-SegCloud is proposed for segmenting cloud images based on a multibranch asymmetric convolution module (MACM) and an attention mechanism. The MACM is composed of asymmetric convolution, depth-separable convolution, and a squeeze-and-excitation module (SEM). The MACM not only enables the network to capture more contextual information in a larger area but can also adaptively adjust the feature channel weights. The attention mechanisms SEM and convolutional block attention module (CBAM) in the network can strengthen useful features for cloud image segmentation. As a result, MA-SegCloud achieves a 96.9% accuracy, 97.0% precision, 97.0% recall, 97.0% F-score, 3.1% error rate, and 94.0% mean intersection-over-union (MIoU) on the Singapore Whole-sky Nychthemeron Image Segmentation (SWINySEG) dataset. Extensive evaluations demonstrate that MA-SegCloud performs favorably against state-of-the-art cloud image segmentation methods.https://www.mdpi.com/2072-4292/14/16/3970ground-based cloud image segmentationconvolutional block attention module (CBAM)squeeze-and-excitation module (SEM)multibranch asymmetric convolution module (MACM)whole sky imager (WSI) |
| spellingShingle | Liwen Zhang Wenhao Wei Bo Qiu Ali Luo Mingru Zhang Xiaotong Li A Novel Ground-Based Cloud Image Segmentation Method Based on a Multibranch Asymmetric Convolution Module and Attention Mechanism Remote Sensing ground-based cloud image segmentation convolutional block attention module (CBAM) squeeze-and-excitation module (SEM) multibranch asymmetric convolution module (MACM) whole sky imager (WSI) |
| title | A Novel Ground-Based Cloud Image Segmentation Method Based on a Multibranch Asymmetric Convolution Module and Attention Mechanism |
| title_full | A Novel Ground-Based Cloud Image Segmentation Method Based on a Multibranch Asymmetric Convolution Module and Attention Mechanism |
| title_fullStr | A Novel Ground-Based Cloud Image Segmentation Method Based on a Multibranch Asymmetric Convolution Module and Attention Mechanism |
| title_full_unstemmed | A Novel Ground-Based Cloud Image Segmentation Method Based on a Multibranch Asymmetric Convolution Module and Attention Mechanism |
| title_short | A Novel Ground-Based Cloud Image Segmentation Method Based on a Multibranch Asymmetric Convolution Module and Attention Mechanism |
| title_sort | novel ground based cloud image segmentation method based on a multibranch asymmetric convolution module and attention mechanism |
| topic | ground-based cloud image segmentation convolutional block attention module (CBAM) squeeze-and-excitation module (SEM) multibranch asymmetric convolution module (MACM) whole sky imager (WSI) |
| url | https://www.mdpi.com/2072-4292/14/16/3970 |
| work_keys_str_mv | AT liwenzhang anovelgroundbasedcloudimagesegmentationmethodbasedonamultibranchasymmetricconvolutionmoduleandattentionmechanism AT wenhaowei anovelgroundbasedcloudimagesegmentationmethodbasedonamultibranchasymmetricconvolutionmoduleandattentionmechanism AT boqiu anovelgroundbasedcloudimagesegmentationmethodbasedonamultibranchasymmetricconvolutionmoduleandattentionmechanism AT aliluo anovelgroundbasedcloudimagesegmentationmethodbasedonamultibranchasymmetricconvolutionmoduleandattentionmechanism AT mingruzhang anovelgroundbasedcloudimagesegmentationmethodbasedonamultibranchasymmetricconvolutionmoduleandattentionmechanism AT xiaotongli anovelgroundbasedcloudimagesegmentationmethodbasedonamultibranchasymmetricconvolutionmoduleandattentionmechanism AT liwenzhang novelgroundbasedcloudimagesegmentationmethodbasedonamultibranchasymmetricconvolutionmoduleandattentionmechanism AT wenhaowei novelgroundbasedcloudimagesegmentationmethodbasedonamultibranchasymmetricconvolutionmoduleandattentionmechanism AT boqiu novelgroundbasedcloudimagesegmentationmethodbasedonamultibranchasymmetricconvolutionmoduleandattentionmechanism AT aliluo novelgroundbasedcloudimagesegmentationmethodbasedonamultibranchasymmetricconvolutionmoduleandattentionmechanism AT mingruzhang novelgroundbasedcloudimagesegmentationmethodbasedonamultibranchasymmetricconvolutionmoduleandattentionmechanism AT xiaotongli novelgroundbasedcloudimagesegmentationmethodbasedonamultibranchasymmetricconvolutionmoduleandattentionmechanism |