面向气溶胶反演的高分四号影像云检测

GF-4 Cloud Detection Algorithm for Aerosol Retrieval

针对在利用卫星影像进行地表信息提取、气溶胶光学厚度反演等工作中云检测不够准确的问题,提出了利用云的空间变化特性、光谱反射率特征阈值和HOT薄云检测联合的方法。通过对大量区域进行直方图统计分析,确定了适合高分四号卫星数据的阈值,进而建立了针对气溶胶反演干扰较大的碎云、云边缘、薄云的检测方法。利用该方法对多幅影像进行云检测,结果表明,该方法能够很好地检测出云像元,特别是对于碎云、云边缘和薄云有很好的识别效果,有效提升了气溶胶产品的可信程度。

There is a problem of inaccurate cloud detection when using satellite imagery for surface information extraction and aerosol optical thickness retrieval.This paper jointed the cloud spatial variability,spectral reflectivity of clouds and the HOT thin cloud detection method,focused on the identification of broken clouds,cloud edges and thin clouds which have great interference with aerosol retrieval,used the histogram statistical analysis of many areas and determined the suitable threshold for GF-4,and then realized cloud detection of GF-4.Using this method,it performed cloud detection on multiple images.The results show that this method can detect cloud pixels very well especially for cloud edges and broken clouds.The method can enhance the credibility of aerosol products effectively.