基于分形网络进化分割和对象特征提取的GF-1卫星数据沙化土地分类识别研究  

The sandy lands identification and classification of GF- 1 based on FNEA and object features

作　　者：李长龙[1] 高志海[1] 吴俊君[1] 孙斌[1] 丁相元 

机构地区：[1]中国林业科学研究院资源信息研究所,北京100091

出　　处：《干旱区资源与环境》2015年第11期152-157,共6页Journal of Arid Land Resources and Environment

基　　金：国家高分辨率对地观测系统重大专项(21-Y30B05-9001-13/15)资助

摘　　要：以高分一号(GF-1)为数据源,以浑善达克沙地为研究区,研究基于GF-1数据的沙化土地分类识别技术。文中的处理范围约为200km*400km,通过J-M距离和最终分类精度来确定每个类别对应的最优分割尺度,分割方法采用的是分形网络进化分割算法(FNEA),通过信息增益比、J48决策树、随机树、标准差和变异系数来确定最优分类对象特征,通过决策树和支持向量机(SVM)结合分类方法形成了半自动化的沙化土地分类识别流程,总体精度达到了85.61%,Kappa系数为0.8295。In this paper,with GF- 1 as data source and the west of the Otindag sandy land as study area,the identification and classification of sandy land based on GF- 1 was studied. The range size was about 200 km ＊400km and the optimal segmentation scale of each category was determined by Jefries- Matusita（ J- M） value and the classification accuracy and the segmentation method was Fractal Net Evolution Approach（ FNEA）. The optimal object features were determined by the information gain ratio,J48 decision tree,random tree,standard deviation and coefficient of variation. They included not only the image spectral features,but also the band ratio,shape index,NDVI and texture features,which enriched the classification feature. The sandy land identification and classification process based on GF- 1 was got by the methods of decision tree and support vector machine（ SVM）. The overall accuracy was 84. 62% and Kappa coefficient was 0. 8182.

关 键 词：GF-1卫星 沙化土地 FNEA 最优分割尺度 最优对象特征 

分 类 号：X144[环境科学与工程—环境科学]