摘要
分类回归树(CART)是一种非参数化的分类与回归方法,在用于遥感影像自动分类时,可方便地应用多源知识,提高分类精度。以延边州试验区土地利用/覆被分类为例,利用分类回归树分析从训练样本中集中发现分类规则,集成遥感影像的光谱特征、纹理特征和辅助地学特征进行分类试验,并与传统的最大似然分类方法进行比较。结果表明,基于CART的决策树分类结果的总精度和Kappa系数分别为90.37%和0.8863,分类精度比MLC监督分类方法有明显提高。
CART was a kind of non- parameter classification and regression method, which used multi - source data to improve classification accuracy when used in automatic classification based on remote sending images. Taking Yanbian area in Jilin Province as a ease study, classification and regression tree (CART) was used to classify the ETM image. Classification tree model integrated spectral, texture and the assistant geographical charaeteristics. The results showed that the accuracy of classification based on the CART was higher than the MLC supervised classification method. And its total classification accuracy was 90.37%, Kappa coefficient was 0. 8863.
出处
《资源开发与市场》
CAS
CSSCI
2011年第2期116-117,130,F0004,F0002,共5页
Resource Development & Market
基金
国家自然科学基金项目"基于多源遥感数据的长白山跨国界地区土地利用/覆盖生态安全格局研究"(编号:40171333)
延边大学"211"重点学科建设项目"东北亚核心区社会与环境长期演化与可持续发展研究"
关键词
纹理特征
光谱特征
CART
决策树
遥感
texture characteristics
spectral characteristics
Classification and Regression Tree (CART)
decision tree
remote sensing