胶东半岛果园TM影像信息的提取决策树方法  

Decision tree classification of orchard information extraction from TM imagery in Jiaodong Peninsula of China

作　　者：于新洋[1,2] 张安定[1] 侯西勇[2] 

机构地区：[1]鲁东大学地理与规划学院,山东烟台264025 [2]中国科学院烟台海岸带研究所/山东省海岸带环境过程重点实验室,山东烟台264003

出　　处：《测绘科学》2012年第4期57-60,共4页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(40801016);中国科学院知识创新工程重要方向项目(kzcx2-yw-224)

摘　　要：本文选取胶东半岛最具代表性的5个果品县(市)为研究区,以Landsat TM影像数据为分类影像,尝试提取果园信息。选用可以"无缝"融入多种辅助信息的决策树分类方法,综合NDVI、地形地貌和缨帽变换等多种辅助信息,利用年内物候变化最大的果园与背景地物的光谱差异,进行果园信息提取;利用SPOT影像以及野外考察资料作为检验样本进行精度验证。表明综合多种辅助信息,利用决策树分类法提取TM影像果园信息可行且准确性较高。Decision tree classification is a kind of classification model which uses certain classification rules to gradually thin the research image.It has been widely used for information extraction from remote sensing images due to its goodness of intuitive and high efficiency.Jiaodong Peninsula is one of the most famous areas in China for the production of fruits;therefore,it is very significant to monitor the distribution of orchards.In this paper,the decision tree classification was used to extract the area of orchard in Jiaodong Peninsula.Specifically,Landsat5 TM image（path120 row034,October24,2005） was available and five most representative cities（Penglai,Longkou,Laizhou,Qixia,Zhaoyuan） were selected as the study area.It turned out that the decision tree classification had satisfactory performance,the classification results were acceptable and could be used as the original inputs for related researches.

关 键 词：TM影像 决策树 果园 信息提取 胶东半岛 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]