The self-adaptive adjustment method of clustering center in multi-spectral remote sensing image classification of land use

Abstract

As one kind of remote sensing images of land use composed by various categories of surface objects difficult to obtain multi-distribution model of class spectral feature, analyzing the spectral characteristics of LU of multispectral RS imagery, this paper presents a self-adaptive adjustment of clustering center method. Depending on the intercepted situation of the cluster centers between different features to conduct split, the sub-centers obtained are as the sub-category features and the cluster centers assemble to characterize category model which is better to deal with the problems of LU category composed by various surface objects and category model not satisfying multivariate normal distribution. As there are much differences between the many centers features in the unit of category area, so the selection of training area and the determinants of rules are easy. The results of experiment indicate that the LU classification accuracy is increased between 4% and 6% with this method. © 2012 IFIP International Federation for Information Processing.