摘要
学习矢量量化(LVQ2)神经网络算法对初值非常敏感,影响遥感图像分类的精度。遗传算法具有很强的全局搜索能力和鲁棒性,能够优化LVQ2神经网络的初始权值向量,在一定程度上降低算法对初值的敏感性。本文采用遗传算法选取LVQ2神经网络的初始权值,并以江苏省扬州地区遥感图像分类为例,通过与标准LVQ神经网络、最大似然法进行比较,结果证明,利用遗传算法的LVQ2神经网络在分类精度上有了一定的提高。
One of the major weak points of Learning Vector Quantization（LVQ） neural network is its sensitivity to the initialization, which affects the remote sensing image classification accuracy. In this paper, Genetic Algorithm（GA） is used to optimize the initial value of LVQ2 neural network to make it less sensitive. The proposed method has been applied to the remote sensing image classification of Yangzhou city, Jiangsu Province, and compared with the standard LVQ neural network and Maximum Likelihood Classifier. The experiment result presents that compared with the general methods, the GA-hased LVQ2 neural network can effectively improve the precision of remote sensing image classification.
出处
《遥感信息》
CSCD
2008年第5期21-24,共4页
Remote Sensing Information
关键词
学习矢量量化
神经网络
遗传算法
遥感图像分类
learning vector quantization
neural network
genetic algorithm
remote sensing lmage classification
