SOM神经网络改进及在遥感图像分类中的应用  

Application and improvement of SOM network in remote sensing image classification

作　　者：任军号[1] 吉沛琦[1] 耿跃[1] 

机构地区：[1]西北工业大学自动化学院,西安710072

出　　处：《计算机应用研究》2011年第3期1170-1172,1182,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(10702065)

摘　　要：针对自组织特征神经网络自身算法的特点和缺陷,采用遗传算法对网络进行改进,形成了基于遗传算法的自组织特征神经网络,并从输入向量、竞争层神经元数量设置和初始权向量设定三方面,结合遥感图像的特性对自组织特征映射网络遥感图像分类的方法进行了改进。将该方法应用于选择西安地区的ETM+卫星遥感图像进行分类实验。结果表明,基于遗传算法的自组织特征映射网络使得遥感图像的分类精度更高,且该算法实现简单,具有一定的工程应用价值。Contrary to the characteristics and weak point of self-organizing maps neural network,this paper shows a SOM which is improved by genetic algorithm,and improves the method to classifying remote sensing image based on self-organizing mapping network through the input vector,selecting the number of competitive layer neurons and the initializing weight vector.Finally,the method is used to classify an ETM＋ satellite remote sensing images of Xi＇an.It is validated that through the improved self-organizing feature map network based on genetic algorithm,classification of remote sensing image have higher accuracy and efficiency than traditional self-organizing feature map network.The SOM based on genetic algorithm is easy to be achieved,and has practical value.

关 键 词：分类 自组织特征映射 神经网络 遗传算法 遥感图像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]