基于高斯混合模型法的国产高分辨率卫星影像云检测  

Gaussian Mixture Model Based Cloud Detection for Chinese High Resolution Satellite Imagery

作　　者：康一飞 潘励[1] 孙明伟[1] 陈奇[2] 王越 

机构地区：[1]武汉大学遥感信息工程学院,湖北武汉430079 [2]中国地质大学(武汉)信息工程学院,湖北武汉430074 [3]武汉市测绘研究院,湖北武汉430022

出　　处：《武汉大学学报（信息科学版）》2017年第6期782-788,共7页Geomatics and Information Science of Wuhan University

基　　金：国家973计划(2012CB719900);国家自然科学基金(41301519)~~

摘　　要：针对资源三号、高分一号等国产高分辨率卫星影像波段少、光谱范围受限的特点,提出一种通过高斯混合模型拟合影像灰度直方图从而自动获取灰度阈值的云检测算法。首先由影像灰度直方图自适应地获取高斯混合模型初始拟合参数,然后依据期望最大原则对初始参数进行调整,最后根据拟合模型中各高斯分量的分布特点,自动确定该波段影像中云与晴空之间的灰度阈值。实验表明,该算法不受限于卫星光谱范围,同时适用于含云和无云影像,检测精度较高,且不需要辅助信息和人工干预,可满足自动化生产的需要。Domestic earth observation satellites, such as ZY-3 and GF-1, have few image bands and limited spectral range. Focusing on those features, this paper outlines a cloud detection algorithm that automatically obtains the grey threshold through histogram fitting by the Gaussian Mixture Model. The algorithm obtains initial parameters automatically from the image histogram; and then it adjusts these parameters by iteration based on the Expectation Maximum principle. It automatically obtains the grey threshold between cloud and clear sky in a chosen image, in line with the distribution features of the components in the Gaussian Mixture Model. Experimental results show that this method has strong advantages, as the range of the spectral bands are not a limit and thus is suitable for both cloudy and cloudless images. In addition, the proposed method needs no auxillary information or manual intervention. It has high computational accuracy and efficiency meeting the requirements for automatic engineering production.

关 键 词：国产高分辨率卫星影像 云检测 高斯混合模型 阈值 

分 类 号：P236[天文地球—摄影测量与遥感]