摘要
以陕西省神木县部分区域为研究区,采用国产高分一号影像数据,结合目视解译成果,研究基于多种特征的CART决策树面向对象分类,对研究区的地物类型进行自动提取。在训练样本一致的前提下,利用面向对象最邻近分类优化特征空间工具间接优化决策树特征空间,并与其分类结果相对比,证明CART决策树面向对象分类是一种高效、高精度的分类方法,尤其适用于大数据量的高分辨率遥感影像地物类型自动提取。
Taking some areas of Shenmu County in Shaanxi Province as the research area,the object-oriented classification of CART decision tree based on multiple features was studied by using domestic No.1 High Resolution Satellite data and combining with visual interpretation results,and the object-oriented classification of the object types in the study area was automatically extracted. With the same training samples,CART decision tree used the nearest neighbor classification tools to optimize feature space,compared with the results of the two classification methods,it showed that the CART decision tree was an efficient and high-precision classification method,especially for the automatic extraction of object type in high resolution remote sensing images with large data volume.
作者
李明
赵英俊
Li Ming;Zhao Yingjun(Beijing Research Institute of Uranium Geology,Beijing 100029)
出处
《资源环境与工程》
2019年第2期251-256,共6页
Resources Environment & Engineering
基金
全国自然资源遥感综合调查与信息系统建设项目(DD20160077)
关键词
高分一号
面向对象
CART决策树
特征空间
No.1 High Resolution Satellite
objected-oriented
CART decision tree classification
Feather Space
