核聚类改进的RBF神经网络遥感影像分类

Improved RBF neural network RS image classification algorithm based on kernel clustering

为提高RBF神经网络遥感影像分类算法的分类效果,本文提出了一种基于核聚类的改进方法。文中采用武汉地区的SPOT影像为实验数据,比较了改进算法与传统基于K均值的RBF神经网络的分类结果,试验表明改进算法的总体精度和Kappa系数均高于传统算法。

The purpose of this article is to enhance the classification performance of RBF neural net- work on remote sensing images by applying an improved algorithm based on kernel clustering. An experi- merit was conducted on：the SPOT image of Wuhan. Comparison of the classification results between the imprdved algorithm with the traditional RBF neural network indicated that the improved method has a higher overall accuracy and Kappa coefficient than the traditional method.