基于Landsat-8影像的大通县土地覆盖分类方法对比研究  

Comparison of methodologies of land cover of Datong county based on Landsat-8 images

作　　者：郅惠[1] 王琪 黄彦丽[1] 李扬 

机构地区：[1]青海省基础地理信息中心,青海西宁810001 [2]青海煤炭地质局测绘工程院,青海西宁810000 [3]青海大学,青海西宁810016

出　　处：《青海大学学报（自然科学版）》2017年第6期83-88,共6页Journal of Qinghai University(Natural Science)

基　　金：基金项目:青海省科学技术厅应用基础研究(2016-ZJ-737)

摘　　要：针对Landsat-8影像在土地覆盖分类中的应用研究,以青海省西宁市大通县为研究区,在对Landsat-8影像进行辐射定标、大气校正、影像裁切等基础上,利用最大似然分类和支持向量机(SVM)分类法,获得两种方法支持下的6种土地覆盖分类结果。经过精度评定和对比分析,结果表明:SVM分类法优于最大似然分类,总体分类精度分别为78.53%,85.64%。同时,Landsat-8 OLI数据相对于TM/ETM+数据,增加的波段新特性有利于土地覆盖分类精度的提高。文中方法适用于Landsat-8影像在土地覆盖分类研究与应用,能够满足大区域土地覆盖分类应用需求。According to the application of land cover classification, this study took the Datong Coun-ty, Xining, Qinghai as the study area. Preprocessing was made on the Landsat - 8 images including the radiometric - calibration, atmospheric correction and image clipping. The classifiers of maximum likelihood and support vector machine were applied to get the two different results of the six land covers. The analysis results showed that the precision of SVM are much better than the precision of Maximum Likelihood, with precision of 78. 5 3 % and 85. 6 4 % . Compared with the data of TM/ETM ＋ , Landsat - 8 0 LI has the new bands which help to promote the classification result of land cover. Meanwhile, the results of this research are meaningful for the relevant land cover classification study of Qinghai Province. The proposed method is suitable for research and application of land cover clas-sification using Landsat - 8 0 LI data and can satisfy the requirements for land cover classification in large regions.

关 键 词：Landsat-8影像 土地覆盖 最大似然 SVM 

分 类 号：P237[天文地球—摄影测量与遥感]