摘要
针对全球变化研究对大洋洲地表覆盖产品的需求,该文以2000年和2010年的Landsat卫星影像为数据源,提出了对大洋洲影像按照月份分组并进行样本采集与规则训练的方法,采用GLC树分类器进行自动分类,经过分类后处理和数据集成,完成了2000年和2010年两期、30m分辨率的大洋洲地表覆盖产品研制工作。利用高分辨率影像、实地采集照片等进行室内精度评定,该大洋洲地表覆盖产品的精度达到90%以上。
As part of Higher resolution Global Land Cover Mapping Project which is initialed by Na- tional High Technology Research and Development Program (" 863" Program), Oceania land cover de- velopment are of great importance. For land cover mapping in Oceania, Landsat ETM+ images in the year 2000 and 2010 are employed and grouped by months. Through sampling and training, land cover are auto- matically classified using GLC decision tree method. The original land cover map are manually modified and updated to the final land cover map in 2000 and 2010. When compared with Google Earth and some pictures in the field, random samples show that the average overall accuracy and Kappa are above 90%.
出处
《测绘科学》
CSCD
北大核心
2016年第11期151-155,共5页
Science of Surveying and Mapping
基金
国家"863"重点项目(2009AA122003)
中国测绘科学研究院基本科研业务费项目(7771507)