摘要
针对传统遥感图像分类算法存在约束条件多、容易陷入局部最优解、分类精度低的缺陷,提出了一种基于自适应差分演化的遥感分类新方法。实验结果表明,基于自适应差分演化的遥感图像分类算法在分类精度上优于传统方法,在收敛速度上优于标准的差分演化分类算法,其分类精度和Kappa系数分别达到了92.66%和0.9017。
A new classification algorithm is proposed for remote sensing imagery based on the self-adapting differential evolution to be waged against the three main disadvantages of traditional classification algorithm for remote sensing image:multiple constraints,easy to fall into local optimal solution,lower classification accuracy.In the new method for supervised classification of multi/hyper-spectral remote sensing image,the globally optimal cluster centers are firstly learned by using the self-adapting differential evolution algorithm,and then the whole remote sensing image can be classified by the cluster centers.The proposed algorithm for classification of remote sensing image is based on the standard differential evolution.The experimental results show that the self-adapting differential evolution clustering algorithm has higher classification accuracy than the traditional classification algorithm of remote sensing image.The classification accuracy and the kappa coefficient can reach 92.66%and 0.9017,which has some practical application value.
作者
吴佳
蔡之华
金晓文
WU Jia;CAI Zhihua;JIN Xiaowen(School of Computer,China University of Geosciences,388 Lumo Road,Wuban 430074.China;School of Environmental Studies,China University of Geosciences,388 Lumo Road,Wuhan 430074,China)
出处
《武汉大学学报(信息科学版)》
EI
CSCD
北大核心
2013年第1期23-26,127,共4页
Geomatics and Information Science of Wuhan University
基金
国家自然科学基金资助项目(61075063)
国家教育部博士点基金资助项目(20090145110007)
中国地质大学(武汉)优秀博士论文基金资助项目
关键词
自适应
差分演化
遥感图像
监督分类
演化算法
self-adapting
differential evolution
RS image
supervised classification
evolutionary algorithm
