中文摘要：

卫星云图是研究天气系统演变规律的重要信息，云层内容从卫星云图中提取出来可以有助于云图分析，减少陆地和海洋信息的干扰。为此采用了模糊C均值聚类算法（FCM）进行云图聚类，该算法具有计算效率高，过程简单的优点，但对初始聚类中心敏感，容易陷入局部最优解。针对此问题，本文将全局性良好的粒子群优化算法（PSO）引入FCM聚类算法，克服了初始聚类中心对全局收敛性的影响。同时，将阴影集理论与该混合算法结合起来，去除聚类过程中的异常值，提高算法的效率。通过红外云图聚类对比实验得出，改进的FCM算法与传统的FCM算法相比，聚类结果图的类间距离增大，类内距离减小，聚类质量有所提高。

英文摘要：

Nephograms can be used to analyze the distribution of the cloud system in a large area, and to study the evolvement rules of weather system. We can analyze the nepbograms without the interference of terrestrial and marine information by extract the clouds content from the nephograms. So fuzzy C-means（FCM） algorithm is used for satellite image clustering. The method is easy to understand, but it always converges to the local infinitesimal values. So PSO was introduced to FCM algorithm, which can find a globally optimal fuzzy segmentation so as to avoid the sensitivities of basic FCM algorithm to initial values. In order to improve the speed of the algorithm, the shadow sets algorithm was combined with FCM, which can remove boundary values and abnormal values. The results show that the clustering effect of the newly-proposed algorithm is better than the one of the basic FCM.