模糊ISODATA聚类算法在声速剖面自动分类中的应用

Clustering of sound speed profile based on fuzzy ISODATA algorithm

依据中国海浅海区30’按月历史统计声速剖面数据，通过归一化处理和Akima差值采样得到梯度剖面，建立起各方区按月归化后的声速剖面分层梯度样本集，并采用模糊ISODATA聚类算法对声速剖面进行聚类分析。通过对分类结果和类内总方差和的分析表明，聚类参数m值在1．1～2．1之间，并以最远邻系统聚类法结果为初始类中心的模糊分类效果较好。应用该方法对海洋中的声速剖面进行自动分类和区划对海洋环境的战术应用意义重大。

A sound speed gradient profile sample database is built through a normalization process and Akima sampling of sound speed profiles which are derived from 30＇× 30＇ latitude-longitude historical statistic data for every month. Fuzzy ISODATA cluster analysis arithmetic is performed to classify the sound speed profiles based on the sample database, and then to calculate sum of variance within the group based on the cluster results. Through analyzing the clustering result and sum of variance within the group, it was shown that the value of clustering parameter m is better between 1. 1-2. 1, and the fuzzy cluster result which takes the farthest neighbour hierarchical cluster result as an initialization cluster center is better. Automatic classifications and provinces of the sound speed profiles are very helpful to the tactical application of marine environments.