摘要
以Matlab神经网络和遗传算法工具箱为平台,用量化共轭梯度法改进标准BP算法,采用GA优化BP网络的隐层神经元数目、初始权重,最后以香格里拉县ETM+图像为数据源,在DEM地形数据辅助下,训练网络使其收敛,仿真结果表明该方法优于最大似然分类法.
In this paper,a new method is presented,in which the neural networks and genetic algorithm toolbox of the Matlab are used as the platform,the conjugate gradient method is adopted to improve the standard BP algorithm,and GA is employed to optimize the BP network to identify the number of hidden layer neurons and the initial weights.As an example,the ETM ＋ remote sensing image of Shangri-La County is classified with this method.The results show that the Kappa coefficient is 0.831 7 and the overall classification accuracy is 84.52%,thus resulting in an improvement of 9.085,as compared with the maximum likelihood classification method.
出处
《西南大学学报（自然科学版）》
CAS
CSCD
北大核心
2010年第7期128-132,共5页
Journal of Southwest University(Natural Science Edition)
基金
国家自然科学基金资助项目(40861009)
关键词
遗传算法
最优化
BP人工神经网络
遥感图像分类
genetic algorithm
optimization
BP neural networks
remote sensing image classification
