中国空间科学技术

中国空间科学技术 ›› 2023, Vol. 43 ›› Issue (1): 1-17.doi: 10.16708/j.cnki.1000-758X.2023.0001

• 综述 •    下一篇

遥感图像云检测方法综述

刘子力，杨家俊，王文静，史振威   

  1. 北京航空航天大学 宇航学院图像处理中心，北京100191
  • 出版日期:2023-02-25 发布日期:2023-01-13

Cloud detection methods for remote sensing images：a survey

LIU Zili，YANG Jiajun，WANG Wenjing，SHI Zhenwei   

  1. Image Processing Center，School of Astronautics，Beihang University，Beijing 100191，China
  • Published:2023-02-25 Online:2023-01-13

PDF (PC)

1726

摘要： 光学遥感图像中云层会对地面信息进行不同程度的遮挡，造成了地表观测信息的模糊和缺失，极大地影响遥感图像的成像质量。因此，对遥感图像中云层覆盖的检测和评估是进一步分析和利用遥感图像信息的基础和关键。通过充分的调研和对比总结，梳理了20世纪90年代以来，国内外基于遥感图像的云检测方法的发展趋势和代表性工作。将基于遥感图像的云检测方法分为三类：基于光谱阈值的方法、基于经典机器学习的方法以及基于深度学习的方法。总结了当前国内外云检测公开数据集，并对比了部分代表性工作的云检测精度。此外，简要梳理了与云检测相关的云雾（霾）检测、云雪检测、云阴影检测以及云去除等方法。对当前云检测相关工作中存在的问题和未来的发展趋势进行了分析和展望。

关键词: 遥感图像, 云检测, 光谱阈值, 统计学习, 深度学习, 综述

Abstract:  The cloud cover in the optical remote sensing images will obscure the ground information in different degrees，which causes the blurring and missing of the surface observation information and greatly affects the imaging quality of remote sensing images.Therefore，the detection and evaluation of cloud cover in remote sensing images are the basis and key to further analyze and utilize remote sensing image information.Through sufficient investigation and summary，the development trend and representative work of cloud detection methods based on remote sensing images at home and abroad since the 1990s were combed.Cloud detection methods based on remote sensing images were divided into three categories：methods based on band threshold，methods based on classical machine learning and methods based on deep learning.Besides，the public datasets at home and abroad used in the related research on cloud detection were summarized，and the accuracy of some representative cloud detection methods was compared.In addition to the vanilla cloud detection method，the cloud/fog（haze）detection，cloud/snow detection，cloud shadow detection and cloud removal methods related to cloud detection were also briefly combed.Based on the review and summary of cloud detection work above，the existing problems and future development trends of cloud detection were analyzed and prospected.

Key words: remote sensing image, cloud detection, band threshold, machine learning, deep learning, survey