中文摘要：

将逐个修改聚类和模糊聚类的多光谱云分类技术应用于2002年6月10日03时锋面气旋云系的识别，并采用同一时次的地面常规观测与其进行了对比分析。结果发现，两种聚类方法对典型锋面气旋云系均有较好的识别能力，分类结果与天气概念模型云层分布情况一致；逐个修改聚类对组间差别较大的情况分类效果较好，而模糊聚类却对性质相近的类别有较好的识别；卫星图像分类结果与地面常规观测比较一致，但在层云、积云、层积云等性质较为相近类别的识别上存在一定差异。

英文摘要：

In this paper a multi-spectral cloud classification technique based on stepwise cluster and fuzzy cluster using GMS-5 imagery data set was applied to recognize frontal cloud systems occurring 03 UTC on 10 June 2002. The results were compared with composite images, routine data and synoptic models, and the reasonability were analyzed. The results show that the typical cloud/surface types can be discriminated from synoptic systems by both classifiers, which agree with clouds distribution of typical synoptic models; most of satellite results are consistent with those of routine data.