Cloud and Cloud Shadow Segmentation for Remote Sensing Imagery Via Filtered Jaccard Loss Function and Parametric Augmentation
Cloud and cloud shadow segmentation are fundamental processes in optical remote sensing image analysis. Current methods for cloud/shadow identification in geospatial imagery are not as accurate as they should, especially in the presence of snow and haze. This article presents a deep learning-based f...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2021-01-01
|
| Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/9394710/ |
| _version_ | 1818620562095734784 |
|---|---|
| author | Sorour Mohajerani Parvaneh Saeedi |
| author_facet | Sorour Mohajerani Parvaneh Saeedi |
| author_sort | Sorour Mohajerani |
| collection | DOAJ |
| description | Cloud and cloud shadow segmentation are fundamental processes in optical remote sensing image analysis. Current methods for cloud/shadow identification in geospatial imagery are not as accurate as they should, especially in the presence of snow and haze. This article presents a deep learning-based framework for the detection of cloud/shadow in Landsat 8 images. Our method benefits from a convolutional neural network, Cloud-Net+ (a modification of our previously proposed Cloud-Net <xref ref-type="bibr" rid="ref1">[1]</xref>) that is trained with a novel loss function [filtered Jaccard loss (FJL)]. The proposed loss function is more sensitive to the absence of foreground objects in an image and penalizes/rewards the predicted mask more accurately than other common loss functions. In addition, a sunlight direction-aware data augmentation technique is developed for the task of cloud shadow detection to extend the generalization ability of the proposed model by expanding existing training sets. The combination of Cloud-Net+, FJL function, and the proposed augmentation algorithm delivers superior results on four public cloud/shadow detection datasets. Our experiments on Pascal VOC dataset exemplifies the applicability and quality of our proposed network and loss function in other computer vision applications. |
| first_indexed | 2024-12-16T17:55:21Z |
| format | Article |
| id | doaj.art-93ec8a8fc6c24412a73bd7a447fa5ff8 |
| institution | Directory Open Access Journal |
| issn | 2151-1535 |
| language | English |
| last_indexed | 2024-12-16T17:55:21Z |
| publishDate | 2021-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| spelling | doaj.art-93ec8a8fc6c24412a73bd7a447fa5ff82022-12-21T22:22:12ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing2151-15352021-01-01144254426610.1109/JSTARS.2021.30707869394710Cloud and Cloud Shadow Segmentation for Remote Sensing Imagery Via Filtered Jaccard Loss Function and Parametric AugmentationSorour Mohajerani0https://orcid.org/0000-0001-6715-5064Parvaneh Saeedi1https://orcid.org/0000-0002-7507-9986School of Engineering Science, Simon Fraser University, Burnaby, BC, CanadaSchool of Engineering Science, Simon Fraser University, Burnaby, BC, CanadaCloud and cloud shadow segmentation are fundamental processes in optical remote sensing image analysis. Current methods for cloud/shadow identification in geospatial imagery are not as accurate as they should, especially in the presence of snow and haze. This article presents a deep learning-based framework for the detection of cloud/shadow in Landsat 8 images. Our method benefits from a convolutional neural network, Cloud-Net+ (a modification of our previously proposed Cloud-Net <xref ref-type="bibr" rid="ref1">[1]</xref>) that is trained with a novel loss function [filtered Jaccard loss (FJL)]. The proposed loss function is more sensitive to the absence of foreground objects in an image and penalizes/rewards the predicted mask more accurately than other common loss functions. In addition, a sunlight direction-aware data augmentation technique is developed for the task of cloud shadow detection to extend the generalization ability of the proposed model by expanding existing training sets. The combination of Cloud-Net+, FJL function, and the proposed augmentation algorithm delivers superior results on four public cloud/shadow detection datasets. Our experiments on Pascal VOC dataset exemplifies the applicability and quality of our proposed network and loss function in other computer vision applications.https://ieeexplore.ieee.org/document/9394710/38-Cloudconvolutional neural network (CNN)cloud detectiondeep learningimage segmentationlandsat 8 |
| spellingShingle | Sorour Mohajerani Parvaneh Saeedi Cloud and Cloud Shadow Segmentation for Remote Sensing Imagery Via Filtered Jaccard Loss Function and Parametric Augmentation IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 38-Cloud convolutional neural network (CNN) cloud detection deep learning image segmentation landsat 8 |
| title | Cloud and Cloud Shadow Segmentation for Remote Sensing Imagery Via Filtered Jaccard Loss Function and Parametric Augmentation |
| title_full | Cloud and Cloud Shadow Segmentation for Remote Sensing Imagery Via Filtered Jaccard Loss Function and Parametric Augmentation |
| title_fullStr | Cloud and Cloud Shadow Segmentation for Remote Sensing Imagery Via Filtered Jaccard Loss Function and Parametric Augmentation |
| title_full_unstemmed | Cloud and Cloud Shadow Segmentation for Remote Sensing Imagery Via Filtered Jaccard Loss Function and Parametric Augmentation |
| title_short | Cloud and Cloud Shadow Segmentation for Remote Sensing Imagery Via Filtered Jaccard Loss Function and Parametric Augmentation |
| title_sort | cloud and cloud shadow segmentation for remote sensing imagery via filtered jaccard loss function and parametric augmentation |
| topic | 38-Cloud convolutional neural network (CNN) cloud detection deep learning image segmentation landsat 8 |
| url | https://ieeexplore.ieee.org/document/9394710/ |
| work_keys_str_mv | AT sorourmohajerani cloudandcloudshadowsegmentationforremotesensingimageryviafilteredjaccardlossfunctionandparametricaugmentation AT parvanehsaeedi cloudandcloudshadowsegmentationforremotesensingimageryviafilteredjaccardlossfunctionandparametricaugmentation |