CDUNet: Cloud Detection UNet for Remote Sensing Imagery

Cloud detection is a key step in the preprocessing of optical satellite remote sensing images. In the existing literature, cloud detection methods are roughly divided into threshold methods and deep-learning methods. Most of the traditional threshold methods are based on the spectral characteristics...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Kai Hu, Dongsheng Zhang, Min Xia
Formato:	article
Lenguaje:	EN
Publicado:	MDPI AG 2021
Materias:	cloud detection cloud shadow multiscale feature fusing deep learning Science Q
Acceso en línea:	https://doaj.org/article/d68877e101ad47b69f9c0d2ef3f6d258
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:d68877e101ad47b69f9c0d2ef3f6d258
record_format	dspace
spelling	oai:doaj.org-article:d68877e101ad47b69f9c0d2ef3f6d2582021-11-25T18:54:00ZCDUNet: Cloud Detection UNet for Remote Sensing Imagery10.3390/rs132245332072-4292https://doaj.org/article/d68877e101ad47b69f9c0d2ef3f6d2582021-11-01T00:00:00Zhttps://www.mdpi.com/2072-4292/13/22/4533https://doaj.org/toc/2072-4292Cloud detection is a key step in the preprocessing of optical satellite remote sensing images. In the existing literature, cloud detection methods are roughly divided into threshold methods and deep-learning methods. Most of the traditional threshold methods are based on the spectral characteristics of clouds, so it is easy to lose the spatial location information in the high-reflection area, resulting in misclassification. Besides, due to the lack of generalization, the traditional deep-learning network also easily loses the details and spatial information if it is directly applied to cloud detection. In order to solve these problems, we propose a deep-learning model, Cloud Detection UNet (CDUNet), for cloud detection. The characteristics of the network are that it can refine the division boundary of the cloud layer and capture its spatial position information. In the proposed model, we introduced a High-frequency Feature Extractor (HFE) and a Multiscale Convolution (MSC) to refine the cloud boundary and predict fragmented clouds. Moreover, in order to improve the accuracy of thin cloud detection, the Spatial Prior Self-Attention (SPSA) mechanism was introduced to establish the cloud spatial position information. Additionally, a dual-attention mechanism is proposed to reduce the proportion of redundant information in the model and improve the overall performance of the model. The experimental results showed that our model can cope with complex cloud cover scenes and has excellent performance on cloud datasets and SPARCS datasets. Its segmentation accuracy is better than the existing methods, which is of great significance for cloud-detection-related work.Kai HuDongsheng ZhangMin XiaMDPI AGarticlecloud detectioncloud shadowmultiscale feature fusingdeep learningScienceQENRemote Sensing, Vol 13, Iss 4533, p 4533 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	cloud detection cloud shadow multiscale feature fusing deep learning Science Q
spellingShingle	cloud detection cloud shadow multiscale feature fusing deep learning Science Q Kai Hu Dongsheng Zhang Min Xia CDUNet: Cloud Detection UNet for Remote Sensing Imagery
description	Cloud detection is a key step in the preprocessing of optical satellite remote sensing images. In the existing literature, cloud detection methods are roughly divided into threshold methods and deep-learning methods. Most of the traditional threshold methods are based on the spectral characteristics of clouds, so it is easy to lose the spatial location information in the high-reflection area, resulting in misclassification. Besides, due to the lack of generalization, the traditional deep-learning network also easily loses the details and spatial information if it is directly applied to cloud detection. In order to solve these problems, we propose a deep-learning model, Cloud Detection UNet (CDUNet), for cloud detection. The characteristics of the network are that it can refine the division boundary of the cloud layer and capture its spatial position information. In the proposed model, we introduced a High-frequency Feature Extractor (HFE) and a Multiscale Convolution (MSC) to refine the cloud boundary and predict fragmented clouds. Moreover, in order to improve the accuracy of thin cloud detection, the Spatial Prior Self-Attention (SPSA) mechanism was introduced to establish the cloud spatial position information. Additionally, a dual-attention mechanism is proposed to reduce the proportion of redundant information in the model and improve the overall performance of the model. The experimental results showed that our model can cope with complex cloud cover scenes and has excellent performance on cloud datasets and SPARCS datasets. Its segmentation accuracy is better than the existing methods, which is of great significance for cloud-detection-related work.
format	article
author	Kai Hu Dongsheng Zhang Min Xia
author_facet	Kai Hu Dongsheng Zhang Min Xia
author_sort	Kai Hu
title	CDUNet: Cloud Detection UNet for Remote Sensing Imagery
title_short	CDUNet: Cloud Detection UNet for Remote Sensing Imagery
title_full	CDUNet: Cloud Detection UNet for Remote Sensing Imagery
title_fullStr	CDUNet: Cloud Detection UNet for Remote Sensing Imagery
title_full_unstemmed	CDUNet: Cloud Detection UNet for Remote Sensing Imagery
title_sort	cdunet: cloud detection unet for remote sensing imagery
publisher	MDPI AG
publishDate	2021
url	https://doaj.org/article/d68877e101ad47b69f9c0d2ef3f6d258
work_keys_str_mv	AT kaihu cdunetclouddetectionunetforremotesensingimagery AT dongshengzhang cdunetclouddetectionunetforremotesensingimagery AT minxia cdunetclouddetectionunetforremotesensingimagery
_version_	1718410621629759488

CDUNet: Cloud Detection UNet for Remote Sensing Imagery

Ejemplares similares