关键词: 模糊C-均值, 二阶模糊聚类, 传递闭包, 分类水平, 距离像

Abstract: Concerning that fuzzy C-means (FCM) algorithm was sensitive to the initial value of cluster centers, a fuzzy clustering method based on Second-Order Fuzzy Clustering (SOFC) was proposed, which took advantage of Transitive Closure (TC) algorithm's non-initialization, the samples were classified according to certain classification level firstly, and then the number of class was selected. The sample means of these classes was used to initialize the FCM algorithms cluster centers. On one hand, it can obtain good initial value of cluster centers; on the other hand, the level of value through the categories can optimize the number of cluster centers and cluster centers, avoiding local optimum and overcoming the consistency of clustering. The algorithm was used to recognize three types of aircraft targets based on onedimensional image data. The experimental results show that the recognition of SOFC algorithm gets obvious improvement than FCM algorithm.

Key words: Fuzzy C-Means (FCM), Second-Order Fuzzy Clustering (SOFC), transitive closure, classification level, range profile