中分辨率卫星数据分类的改进迭代聚类算法

在动态聚类算法的理论基础上,提出了一种经过改进的迭代聚类方法。该方法通过VC＋＋程序编译加以实现,在分类的过程中通过判定类间距离阈值动态分类,无需设定初始聚类中心,减少了分类前需要设定的初始参数,保证了分类结果更加客观,并且分类过程简便易操作。为了验证其实用性,本文选取了辽宁中部一区域作为试验区,应用TERRA卫星MOD IS和FY3A卫星MERSI 250m中分辨率数据进行分类,并同时与ISODA-TA法的分类结果进行比较。结果表明,使用该方法得到的分类精度好于ISODATA法,可以用于遥感数据的分类。

An Improved Iterative Clustering Algorithm used in the classification of moderate resolution satellite data

Based on the theory of dynamical clustering, an improved iterative clustering algorithm was brought forward. This iterative clustering algorithm realized on the Microsoft Visual C++ platform by judging the threshold of distance between classes, and had following advantages, such as did not need to predefine initial clustering centers, reduced number of initial parameters to be decided before the classifying, guaranteed more realistic classifying results and 'made the clustering process easier. In order to validate the practicability of this algorithm, the South Liaoning Province was chosen as a study area, and the moderate resolution satellite data of 250m TERRA MODIS and 250m MERSI of FY3A satellite were used to compared with ISODATA algorithm. The results indicated that the classifying precision of new a-rithmetic was better, and could be used in classifying remote sensing data.