Abstract: A Growing Hierarchical Self-Organizing Map（GHSOM） based remote sensing image classification method is presented.Its theory and algorithm are introduced firstly，and then it is successfully applied to remote sensing image classification.Experimental results show that GHSOM solves classification confusion problem of SOM，and greatly improves classification accuracy and efficiency for remote sensing data.GHSOM method is a new efficient unsupervised remote sensing image classification method.

Key words: Self-Organizing Map（SOM）, image segmentation, Growing Hierarchical SOM

摘要： 提出一种基于自组织增长分级神经网络（Growing Hierarchical Self-Organizing Map ，GHSOM）的遥感图像分类方法。首先详细分析了GHSOM方法的基本原理和算法，然后成功将其应用于遥感图像分类。实验结果表明了GHSOM通过分级的分类方法有效解决了SOM分类中的混分问题，大大提高了分类精度和效率，是一种新的有效的无监督遥感图像分类方法。

关键词: 自组织神经网络, 图像分割, 自组织增长分级神经网络