设计完成和比较了自组织网络的几种算法在遥感土地覆盖分类中的应用,结果表明非监督和监督学习相结合方法进行遥感土地覆盖分类,各算法在分类性能上无显着差异,因此可采用算法较简单的简单竞争学习网络,根据最邻近原则进行非参数分类。

In this peaper, the implement and comparison of different self organizing learn ing algorithm in landcover classification of Landsat TM imagery it is found that, with the combination of unsupervised and supervised learning method and the ne a rest neighbour principle, these algorithms have no significant difference in cla ssification accuracy. The study result shows that the self organizing network i s an another method to classify the landcover type in remote sensing imagery by combining the unsupervised and supervised learning phase with the nearest neighb our principle. Because of the simplicity of the Simple Competivite Learning, the self organizing network can use the Simple Competivite Learning algorithm in re motely sensed data classification.