摘要
该研究以Matlab为平台,应用自组织特征映射网络模型对研究区的遥感影像数据进行土地使用类型的分类。结果表明:在设计神经网络模型时,输出层节点数的确定不仅与分类数相关,同时也需要考虑研究区数据的模数,输出层节点数正确与否很大程度上制约着研究结果的精度。研究证明,自组织特征映射网络通过神经元之间的竞争能模拟大脑神经系统中的"近兴奋远抑制"功能使得该网络的收敛性更好,其分类精度较高,而且该神经网络不需要学习样本使其应用更加简单。因此自组织特征映射网络在遥感分类中有着很好的应用前景。
Sell-organizing ieature map (SOFM) neural network was applied to classify remote sensing images of Haidian District,Beijing with regard to land-use types.The results show that the nodes of the output layer have relations not only with the number of clusters but also with the modulus of the study area,and the nodes of the output layer have a marked effect on the accuracy of the classification.The results indicate that SOFM can simulate the function of brain neural network through competition among neurons without learning samples, and its convergence is good and the classification accuracy is high.SOFM has a promising application prospect in image classification.
出处
《北京林业大学学报》
CAS
CSCD
北大核心
2008年第S1期73-77,共5页
Journal of Beijing Forestry University
基金
农业科技成果转化资金项目(05EFN217100428)
关键词
自组织特征映射
竞争
拓扑保形性
图像分类
Self-organizing feature competition (SOFM)
topological shape preserving
image classification
