决策树方法从SPOT-5卫星影像中自动提取水体信息研究  

Study on the automatic extraction of water body information from SPOT-5 images using decision tree algorithm

作　　者：邓劲松[1] 王珂[1] 李君[2] 董云奇 

机构地区：[1]浙江大学农业遥感信息技术应用研究所,浙江杭州310029 [2]浙江大学环境工程系,浙江杭州310027 [3]浙江省桐乡市土地资源局,浙江桐乡310045

出　　处：《浙江大学学报（农业与生命科学版）》2005年第2期171-174,共4页Journal of Zhejiang University:Agriculture and Life Sciences

基　　金：国家自然科学基金(40201021);浙江省科技厅(2004c33089)资助课题.

摘　　要：分析了SPOT-5影像中水体及其它主要地物的光谱特性及波段间的关系,由此探讨它们在光谱特征上的可分性.研究发现,不同波段之间只有水体具有B3(green)>B4(SW)且B2(red)>B1(IR)的特殊关系,同时在短波红外波段(SW)上,水体与其它地物亮度值差异明显,可以通过设置阀值加以区分.根据以上分析,建立了决策树模型,在各节点设计不同的分类器,进行水体信息的提取并对提取结果进行了精度评价.结果表明,该方法的总体提取效果较好,其提取精度与通常的监督分类方法相比有了较大的提高,只是在水体和其它地物交界处有误判现象.Extraction of water information from remotely sensed images is significant for surveying, planning and protecting water resources. In this paper, taking Tongxiang county of Jiaxing as a case study area, SPOT-5 image collected on Aug, 2002 was used. The approach of extraction was discussed. Firstly, The mechanism and characteristics of water body and other objects in SPOT-5 imagery was analyzed to find the possibility of extracting farmland from the background. The results show that water body is distinguishable from background in SW band and hase special relationship B3 (green)>B4 (SW) and B2 (red)>B1 (IR) between bands. Secondly, based on the analysis, a simple model of decision tree was applied to extracting water information. Finally, the results are checked by visual and statistical accuracy assessment. The results suggest that the model of decision tree was simple and effective and the precision of this approach was much higher than that of the supervised classification. However, some pixels in the neighborhood between water bodies and mulberry are judged by mistakes.

关 键 词：SPOT-5卫星影像 水体 信息提取 决策树 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]