Abstract
ABSTRACT Cloud detection is one of the important tasks for remote sensing image preprocessing. In this work, multiple infrared observation channels with high time resolution are utilized to extract cloud from geostationary satellite imagery. Compared with detected cloud mask from polar satellite data with the high spatial resolution, the detection from geostationary satellite data has high timeliness and practicability, but it is challenging for the detailed segmentation ability of the network. Due to the fact that some small clouds only occupy a few pixels in geostationary satellite images with low spatial resolution, the existing network has poor segmentation ability for these small targets. To tackle this problem, a new neural network named U-High Resolution Network (U-HRNet) is proposed for cloud detection. Combing both multi-scale feature extraction using a High Resolution Network (HRNet) architecture and merging shallow information and deep information via the skip connection (SC) of a U-Net structure, the proposed U-HRNet produces strong high-resolution representations for accurate detection of details on cloud segmentation. The accuracy of the method is evaluated by manually labelled ground truth against different methods using objective evaluation indices. The results proved the proposed U-HRNet performs well on FengYun-4A (FY-4A) images and can effectively detect incorrect areas of the cloud mask products of the National Satellite Meteorological Centre (NSMC), depicting outperformance over existing methods.
Published Version
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