Abstract
The presence of clouds greatly reduces the ground information of high-resolution satellite data. In order to improve the utilization of high-resolution satellite data, this article presents a cloud removal method based on deep learning. This is the first end-to-end architecture that has great potential to detect and remove clouds from high-resolution satellite data. For cloud detection, a convolution neural network (CNN) architecture is used to detect them. For cloud removal, the content generation network, the texture generation network, and the spectrum generation network based on traditional CNN are proposed. The proposed CNN architecture can use multisource data (content, texture, and spectral) as an input of the unified framework. The results of both the simulated and real image experiments demonstrate that the proposed method is robust and can effectively remove thick clouds, thin clouds, and cloud shadows. In addition, compared with some existing methods, the proposed method can recover land cover information accurately.
Highlights
Z I YUAN-3 (ZY-3) satellite simultaneously offers mapping and resource survey functions, providing reliable data sources
3) In the cloud removal stage, we present the content generation network, the texture generation network, and the spectrum generation network based on traditional convolution neural network (CNN) to improve the clouds recovering accuracy and consistency
Where Goutput is the content-texture-spectral CNN output feature map, Ginput 1 is the feature map produced by texture CNN, Ginput 2 is the feature map produced by spectral CNN, C(Ginput 1, Ginput 2) stands for the entire input feature maps obtained by concatenating Ginput 1 to Ginput 2, and V (C(Ginput 1, Ginput 2) denotes the feature map produced by convolutional kernel taking C(Ginput 1, Ginput 2)as input data
Summary
Z I YUAN-3 (ZY-3) satellite simultaneously offers mapping and resource survey functions, providing reliable data sources. Multitemporal-based methods hold the assumption that land-cover representation does not greatly change between cloud-contained images and cloud-free images They only utilized contaminated images and cloudfree images, and the latent temporal correlations in multitemporal data were not utilized. The powerful nonlinear expression ability of deep learning can be introduced for recovering degraded remote sensing images [21] It makes the global optimization training parameters greatly reduced and has a wide range of applications in the field of image recognition and recovery [22]. This article introduces the cloud removal method of multitemporal ZY-3 remote sensing images based on deep learning. 1) We present the cloud removal method of a multitemporal ZY-3 satellite image based on deep learning.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
