高分辨率影像的干旱区植被信息提取方法  

Methods to extract vegetation information based on the high-resolution image for arid region

作　　者：何磊[1] 王超[1] 别强[1] 赵传燕[2] 

机构地区：[1]兰州大学遥感与地理信息系统研究所,兰州730000 [2]兰州大学草地农业系统国家重点实验室,兰州730000

出　　处：《干旱区资源与环境》2012年第11期132-138,共7页Journal of Arid Land Resources and Environment

基　　金：国家自然科学基金重点项目(91025015);国家环境保护公益性资助项目(NEPCP 20809098);中国-联合国合作非洲水行动项目(2010DFA32850)资助

摘　　要：基于黑河下游额济纳旗地区的Quickbird影像,采用决策树(Decision Tree)、人工神经网络(Artifi-cial neural net,ANN)及支持向量机(Support Vector Machine,SVM)方法对干旱区植被信息进行提取。对三种方法的精度进行评价,结果显示:决策树分类得到的结果零碎,总体分类精度为84.87%;ANN法较决策树方法适宜度高,总体分类精度为91.87%;纹理信息辅助的SVM法取得效果最好,总体分类精度可达96.53%。试验中发现使用高分辨率影像提取干旱区植被种类信息时,大窗口的纹理特征辅助效果较好,但是分类结果的边界出现失常,随着纹理窗口越大,失常的范围也越大。In this study,we selected Ejina oasis in the lower reaches of Heihe River as a study area.Three methods such as decision tree,Artificial Neural Network（ANN） and Support Vector Machine（SVM） were applied to classify land use types in two typical fields based on the Quickbird image data.The study results showed that the classification was fragmentary by using decision tree method with the overall accuracy of 84.87% in experimental zone 1.The method of ANN was suitable in the study area with the overall accuracy of 91.87%.SVM with the auxiliary of texture information was the best way to classify land use types with the overall accuracy of 96.53%. The study found that the classification was preferable when using the high-resolution image with assistance of texture at large window scale in the extraction of vegetation types in arid areas.But the boundary of the classification would greatly appear disorders with the bigger window of texture calculations.

关 键 词：干旱区 高分辨率影像 纹理特征 决策树 人工神经网络 支持向量机 

分 类 号：Q948.15[生物学—植物学]