中文摘要：

介绍一种基于图像熵和自动分割技术的卫星云图阂值分类方法，重点解决普通阈值法中阈值不确定性的问题。采用自动分割方法进行云图的划分，结合图像纹理特征信息，通过统计分析来确定具体的阈值。将该方法应用于2007年7月上旬淮河流域的暴雨过程，并与美国GOES卫星分层阈值进行类比，分析了两种阈值与地面实况降水间的关系。结果表明，该方法能够在卫星估算降水过程中增强对不同类型降水的识别能力，是一种有效的卫星云图分类方法。

英文摘要：

The problem of thresholds＇ uncertainty in satellite cloud imagery classification has been resolved using a technique based on the information of image entropy. Based on the Kapur technique of image entropy, the cloud imageries are classified into several kinds automatically. And the nine thresh- olds selected as the classification results are analyzed. With the nine thresholds, the cloudy imageries are classified into ten cloud layers. And the cloud classification method is realized. During the storm rainfall of the first ten days of July 2007 in Huaihe River Basin, the method is applied and the results are com- pared with the enhancement technology of infrared image in GOES User＇s Guide. The results show that the method is useful for analyzing ground precipitation based on satellite image.