基于动态聚类的MODIS云检测算法  

Cloud Detection of MODIS Satellite Images Based on Dynamical Cluster

作　　者：刘志刚[1] 李元祥[2] 黄峰[1] 

机构地区：[1]解放军理工大学气象学院 [2]上海交通大学空天科学技术研究院

出　　处：《遥感信息》2007年第4期33-35,75,I0004,共5页Remote Sensing Information

基　　金：国家自然科学基金资助项目(40575010)

摘　　要：利用多光谱阈值法进行云与下垫面的初始化分类,然后应用基于最小距离和分类矩阵的动态聚类算法,对MODIS云图实现云检测,并分离出积雨云、卷云、其他高云、中云、低云。针对我国东南沿海地区的试验结果证明了算法的有效性。Several classes of cloud and scene domain were first initialized by using multiple spectrum thresholds. Based on mini- mum distance and classification matrix, a dynamical clustering algorithm was used to successfully detect the cloud from MODIS satellite images. At the same time, five kinds of cloud which are cumulonimbus, cirrus, other high cloud, middle cloud and low cloud were obtained separately. Finally, experimental results of Chinese southeast sea coaster region show that the cloud detection algorithm is effective.

关 键 词：云检测 动态聚类 分类矩阵 多光谱阈值 

分 类 号：P237.9[天文地球—摄影测量与遥感]