Cloud and Snow Identification Based on DeepLab V3+ and CRF Combined Model for GF-1 WFV Images
Cloud and snow identification in remote sensing images is critical for snow mapping and snow hydrology research. Aimed at the problem that the semantic segmentation model is prone to producing blurred boundaries, slicing traces and isolated small patches for cloud and snow identification in high-res...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-09-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/14/19/4880 |
| _version_ | 1797477139220529152 |
|---|---|
| author | Zuo Wang Boyang Fan Zhengyang Tu Hu Li Donghua Chen |
| author_facet | Zuo Wang Boyang Fan Zhengyang Tu Hu Li Donghua Chen |
| author_sort | Zuo Wang |
| collection | DOAJ |
| description | Cloud and snow identification in remote sensing images is critical for snow mapping and snow hydrology research. Aimed at the problem that the semantic segmentation model is prone to producing blurred boundaries, slicing traces and isolated small patches for cloud and snow identification in high-resolution remote sensing images, the feasibility of combining DeepLab v3+ and conditional random field (CRF) models for cloud and snow identification based on GF-1 WFV images is studied. For GF-1 WFV images, the model training and testing experiments under the conditions of different sample numbers, sample sizes and loss functions are compared. The results show that, firstly, when the number of samples is 10,000, the sample size is 256 × 256, and the loss function is the Focal function, the model accuracy is the optimal and the Mean Intersection over Union (MIoU) and the Mean Pixel Accuracy (MPA) reach 0.816 and 0.918, respectively. Secondly, after post-processing with the CRF model, the MIoU and the MPA are improved to 0.836 and 0.941, respectively, compared with those without post-processing. Moreover, the misclassifications such as blurred boundaries, slicing traces and isolated small patches are significantly reduced, which indicates that the combination of the DeepLab v3+ and CRF models has high accuracy and strong feasibility for cloud and snow identification in high-resolution remote sensing images. The conclusions can provide a reference for high-resolution snow mapping and hydrology applications using deep learning models. |
| first_indexed | 2024-03-09T21:14:20Z |
| format | Article |
| id | doaj.art-cafaab50c86f402c8ecdee4963d06ceb |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-09T21:14:20Z |
| publishDate | 2022-09-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-cafaab50c86f402c8ecdee4963d06ceb2023-11-23T21:40:14ZengMDPI AGRemote Sensing2072-42922022-09-011419488010.3390/rs14194880Cloud and Snow Identification Based on DeepLab V3+ and CRF Combined Model for GF-1 WFV ImagesZuo Wang0Boyang Fan1Zhengyang Tu2Hu Li3Donghua Chen4School of Geography and Tourism, Anhui Normal University, Wuhu 241002, ChinaSchool of Geography and Tourism, Anhui Normal University, Wuhu 241002, ChinaSchool of Geography and Tourism, Anhui Normal University, Wuhu 241002, ChinaSchool of Geography and Tourism, Anhui Normal University, Wuhu 241002, ChinaSchool of Geography and Tourism, Anhui Normal University, Wuhu 241002, ChinaCloud and snow identification in remote sensing images is critical for snow mapping and snow hydrology research. Aimed at the problem that the semantic segmentation model is prone to producing blurred boundaries, slicing traces and isolated small patches for cloud and snow identification in high-resolution remote sensing images, the feasibility of combining DeepLab v3+ and conditional random field (CRF) models for cloud and snow identification based on GF-1 WFV images is studied. For GF-1 WFV images, the model training and testing experiments under the conditions of different sample numbers, sample sizes and loss functions are compared. The results show that, firstly, when the number of samples is 10,000, the sample size is 256 × 256, and the loss function is the Focal function, the model accuracy is the optimal and the Mean Intersection over Union (MIoU) and the Mean Pixel Accuracy (MPA) reach 0.816 and 0.918, respectively. Secondly, after post-processing with the CRF model, the MIoU and the MPA are improved to 0.836 and 0.941, respectively, compared with those without post-processing. Moreover, the misclassifications such as blurred boundaries, slicing traces and isolated small patches are significantly reduced, which indicates that the combination of the DeepLab v3+ and CRF models has high accuracy and strong feasibility for cloud and snow identification in high-resolution remote sensing images. The conclusions can provide a reference for high-resolution snow mapping and hydrology applications using deep learning models.https://www.mdpi.com/2072-4292/14/19/4880cloud and snow identificationsemantic segmentationdeep neural networkDeepLab v3+conditional random fieldGF-1 image |
| spellingShingle | Zuo Wang Boyang Fan Zhengyang Tu Hu Li Donghua Chen Cloud and Snow Identification Based on DeepLab V3+ and CRF Combined Model for GF-1 WFV Images Remote Sensing cloud and snow identification semantic segmentation deep neural network DeepLab v3+ conditional random field GF-1 image |
| title | Cloud and Snow Identification Based on DeepLab V3+ and CRF Combined Model for GF-1 WFV Images |
| title_full | Cloud and Snow Identification Based on DeepLab V3+ and CRF Combined Model for GF-1 WFV Images |
| title_fullStr | Cloud and Snow Identification Based on DeepLab V3+ and CRF Combined Model for GF-1 WFV Images |
| title_full_unstemmed | Cloud and Snow Identification Based on DeepLab V3+ and CRF Combined Model for GF-1 WFV Images |
| title_short | Cloud and Snow Identification Based on DeepLab V3+ and CRF Combined Model for GF-1 WFV Images |
| title_sort | cloud and snow identification based on deeplab v3 and crf combined model for gf 1 wfv images |
| topic | cloud and snow identification semantic segmentation deep neural network DeepLab v3+ conditional random field GF-1 image |
| url | https://www.mdpi.com/2072-4292/14/19/4880 |
| work_keys_str_mv | AT zuowang cloudandsnowidentificationbasedondeeplabv3andcrfcombinedmodelforgf1wfvimages AT boyangfan cloudandsnowidentificationbasedondeeplabv3andcrfcombinedmodelforgf1wfvimages AT zhengyangtu cloudandsnowidentificationbasedondeeplabv3andcrfcombinedmodelforgf1wfvimages AT huli cloudandsnowidentificationbasedondeeplabv3andcrfcombinedmodelforgf1wfvimages AT donghuachen cloudandsnowidentificationbasedondeeplabv3andcrfcombinedmodelforgf1wfvimages |