基于决策树系统的遥感植被分类技术  

Vegetation classification technology of hyperspectral remote sensing based on decision tree tool

作　　者：何诚[1] 董志海[2,3] 张思玉[1] 冯仲科[2] 郭晓晓[2] 

机构地区：[1]南京森林警察学院,南京210023 [2]北京林业大学测绘与3S研究中心,北京100083 [3]北京市测绘设计研究院,北京100038

出　　处：《测绘科学》2014年第1期83-86,共4页Science of Surveying and Mapping

基　　金：国家863计划(2009AA12Z327;2008AA121305-4);国家科技支撑计划项目(2012BAH34B01);北京市自然科学基金项目(09D0297);国家自然科学基金(30872038)

摘　　要：为了将宏观模糊的农业种植结构定量、准确、直观地反映出来,本文运用MODIS植被指数对作物生长规律分析,并与决策树分类系统相结合,成功提取了河北省黑龙港区域的冬小麦、棉花、夏玉米、花生、果树和蔬菜的分布信息。研究结果表明:当采用基于决策树系统的TM影像与MODIS_EVI影像相结合的分类方法精度较高,总体精度可达91.3%,蔬菜、小麦、棉花、玉米等4种作物较传统影像监督分类的结果分别提高了13.8%、2.0%、1.3%、20.5%。In order to express the macroscopic and fuzzy crop planting structure with quantitative, ac- curate and intuitive description, this paper used the method of integrating the multi-temporal and full-cov- ered MODIS and medium resolution sample TM data, through the human-computer interaction means, which is the combination of analysis of MODIS vegetation indices on the crop growth laws and decision tree classification system. At last, the distribution information of winter wheat, cotton, maize, peanuts, fruit and vegetables which in Heilonggang was extracted successfully. The results showed that the accuracy of classification used by combination of TM images based on decision tree system and MODIS _ EVI images could be up to 91.5%, and comparing with traditional image supervised classification results, those of veg- etables, wheat, cotton, corn and maize could increase by 13.8%, 2.0%, 1.3% and 20.5% respectively.

关 键 词：决策树模型 MODIS TM样本 结构现状 监测 

分 类 号：TP753[自动化与计算机技术—检测技术与自动化装置] P237.9[自动化与计算机技术—控制科学与工程]