决策树在高分辨率影像目标为小地物的分类应用研究  

On the application of the decision tree to target in the study of small object classification by using high-resolution image

作　　者：张妤琳[1,2] 赵宇[2] 刘晓南[2] 

机构地区：[1]韶关学院英东农业科学与工程学院,广东韶关512005 [2]广东省生态环境与土壤研究所,广东广州510650

出　　处：《韶关学院学报》2014年第6期56-59,共4页Journal of Shaoguan University

基　　金：韶关学院科研项目(韶学院[2013]1号文);韶关学院大学生创新训练项目(Sycxcy2013-014);广东省大学生创新训练项目(1057613-007)

摘　　要：高分辨率遥感影像的应用越来越多,但其高昂的成本让一般项目望而却步.应用软件从Google Earth上下载已成为获取高空间分辨率影像的有效途径,但因无光谱信息,解译局限较大.选用C4.5算法的决策树方法,对目标为水塘研究的广州市天河区的下载的快鸟数据进行解译,与最大似然分类法和面向对象分类法相比较.结果表明:决策树分类法的分类精度和kappa系数均较高,能利用多源数据,结构简单直观,易于表达和应用;提取小目标地物更有效,数据量相对小,速度较快.The applications of high resolution remote sensing image in many fields become increasingly helptul, but the cost is formidable. Downloading high spatial resolution image from the Google Earth by applications of software has become an effective way. However, because there is no spectral information in the downloaded data, so the interpretation is limited. On the basis of mainstream remote sensing method, the paper selected the decision tree method by C4.5 algorithm, to interpret quick bird image of Tianhe district of Guangzhou city, which projected target to the pond, and then compared results of the maximum likelihood classification method and object-oriented classification. Analysis showed that the decision tree classification method had the following results： both the overall accuracy and Kappa coefficients were higher; it could reduce consumption of knowledge establishment and practice; it could make use of multi-source data; it was simple and intuitive structure, easy to express and application; the extracted small target features was more effectively, data volume was small relatively; it run fast.

关 键 词：Quick Bird遥感数据 决策树 塘 天河区 

分 类 号：P237.3[天文地球—摄影测量与遥感]