基于ZY-3卫星影像多层分割的村镇用地决策树分类  

Decision tree classification of land uses in villages and towns based on the multi-level segmentaion of ZY-3 images

作　　者：江振蓝[1] 潘辉[2] 

机构地区：[1]闽江学院地理科学系,福建福州350108 [2]闽江学院科研处,福建福州350108

出　　处：《闽江学院学报》2016年第2期115-123,共9页Journal of Minjiang University

基　　金：福建省教育厅项目(JA12267);福州市科技计划项目(2014S130);闽江学院科技项目(YKQ13005);福建省测绘地理信息局项目(2014M02)

摘　　要：以资源三号(ZY-3)卫星影像为研究对象,采用多层次影像分割的决策树分类方法对长乐市鹤上镇的村镇用地信息进行提取.首先,在ENVI软件的支持下,通过反复分割试验,确定各村镇用地的最优分割尺度和合并尺度,并对影像进行多层次分割;然后,通过村镇用地对象的光谱、纹理、形状特征分析,筛选出各地类相互间区别明显的对象特征,建立村镇用地分类的决策树规则,并进行村镇用地信息的提取.研究表明,较传统的分类方法,该方法分类的精度有很大提高(总体精度为92.6%,Kappa系数为0.91),可为高分辨率遥感影像的村镇用地信息提取提供参考.A decision tree classification based on multi-level segmentation of ZY-3 images was conducted on land use in Heshang Town of Fujian Province. At first,the images were segemented for every land use in the study area by using the optimal segmentation and merger scale identified through a trialand-error approach on the basis of ENVI. Then,an analysis was made on the spectral,texture and shape features of land uses in the villages and towns to abstract characters that identify themselves each distinctively and set up decision tree rules accordingly. At last,land uses in villages and towns were abstracted by using the decision tree method. Compared with traditional approach of pixel-based classification and single-level object-oriented classification,the approach introduced in this research features greater precision（ with overall accuracy being 92. 6% and Kappa efficience 0. 91） and applicability and can be used as reference for land use abstraction in similar areas from high-resolution RS images.

关 键 词：ZY-3影像 村镇用地 多层次影像分割 决策树 面向对象 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]