摘要
植被覆盖是国家开展的地理国情普查工作的重要组成部分。文章以高分辨率遥感影像为基础,完成影像预处理工作后,获取影像的光谱特征、归一化植被指数与纹理特征,由三种特征共同组成待分类影像,通过人工交互为每种植被选取样本,利用二叉决策树方法创建植被的分类规则,从而完成自动分类。将分类结果与地理国情普查的地表覆盖成果相互验证,同时对比实验表明分类实验总分类精度与Kappa系数均高于最大似然法的分类精度与Kappa系数,证明获取的二叉树的有效性。后续可利用二叉树快速更新,更新结果作为地理国情更新参考数据。
Vegetation coverage is an important part of the national general survey of geographical conditions.This paper takes high resolution images as the data source.After pre-processing images,three kinds of characters of images,including spectral features,NDVI and texture features,are obtained to form images to be classified,classification rules are established subsequently with the method of the binary decision tree to complete automatic classification after sample selection through human interaction.Classfication results and survey results of geographical conditions are tested each other,contrast experiments show that overall accuracy and kappa coefficients are higher than that of maximum likelihood method,which indicates effectiveness of the binary tree.The obtained binary tree can be used for rapid update of geographical conditions.Classification results are classifyed according to standards of geographic national conditions census,results of which also helps to evaluate the classification accuracy.
作者
陈超
赫春晓
CHEN Chao;HE Chun-xiao(Provincial Geomatics Center of Jiangsu,Nanjing Jiangsu 210013,China;Jiangsu Provincial Research Institute of Surveying&Mapping,Nanjing Jiangsu 210013,China)
出处
《现代测绘》
2019年第5期28-31,共4页
Modern Surveying and Mapping
基金
基于纹理信息与决策树的植被分类方法研究(JSCHKY201421).
关键词
纹理特征
决策树
植被分类
地理国情普查
texture
decision tree
vegetation classification
geographic national conditions census
