基于指数的遥感影像决策树分类方法

发布时间：2018-11-27 19:52

【摘要】：本研究以哈尔滨市为研究区,采用Landsat-8多光谱影像为数据源,计算出水体指数、植被指数、建筑指数和土壤指数等共20个地物光谱指数;基于指数的决策树分类方法提取土地覆被类型,并对分类结果进行精度验证,对比分析不同指数对分类精度的影响。经过精度对比、筛选五组数据作为变量用决策树方法进行分类,并与于单纯地物光谱的(普通)决策树分类结果和最大似然法相比较。结果显示:五组基于指数的决策树分类结果都比普通决策树分类和传统的最大似然法分类精度高,分类精度最高的一组与以上两种分类结果相比,总精度分别提高了2.59%和9.55%,Kappa系数分别提高了0.08和0.15。本研究呈现了基于指数的决策树分类方法在土地利用信息提取中的优势,为更好协调哈尔滨市土地利用与城市扩展提供研究依据。
[Abstract]:In this study, 20 spectral indices of water body, vegetation index, building index and soil index were calculated by using Landsat-8 multi-spectral image as data source and Harbin city as study area. The decision tree classification method based on index is used to extract land cover types, and the accuracy of classification results is verified, and the effects of different indices on classification accuracy are compared and analyzed. After precision comparison, five groups of data were selected as variables and classified by decision tree method, and compared with the classification results of the (common) decision tree and the maximum likelihood method. The results show that the classification accuracy of the five groups of decision trees based on index is higher than that of the ordinary decision trees and the traditional maximum likelihood method. The group with the highest classification accuracy is compared with the above two kinds of classification results. The total precision was increased by 2.59% and 9.55%, respectively, and the Kappa coefficient was increased by 0.08 and 0.15, respectively. This study presents the advantage of decision tree classification method based on index in extracting land use information, which provides a basis for better coordination of land use and urban expansion in Harbin.
【作者单位】：哈尔滨师范大学地理科学学院;黑龙江省普通高等学校地理环境遥感监测重点实验室;
【基金】：国家自然科学基金(41571199) 黑龙江省自然科学基金项目(ZD201308);黑龙江省自然科学基金青年项目(QC2016050)
【分类号】：P237

【相似文献】