摘要
GlobeLand30的发布为全球生态系统评价、地理国情监测、环境保护等研究提供了高分辨率的基础地理信息,其质量的优劣直接关系到相关应用及决策的可靠性。本文以江西省为实验区,通过引入景观形状指数对传统的精度评定方法进行了改进,实现对Globeland30遥感产品的精度评定。以传统的概率统计理论计算总体样本量,采取内曼分配与按面积比例分配相结合的方式进行样本量的逐层分配,样本布设时在分层布设的基础上引入景观形状指数,以提高样本的典型性和评估结果的可靠性。利用混淆矩阵,计算出江西省地表覆盖产品的总体精度为81.94%,Kappa系数为0.7005,这表明实验区域GlobeLand30地表覆盖遥感产品质量可靠。
The publication of GlobeLand30 provides the highest resolution basic information for the global ecosystem evaluation,geographical situation monitoring and environmental protection research,and the reliability of its quality is directly related to the application and decision making.In this paper,the traditional method of precision evaluation is improved by introducing landscape shape index in Jiangxi Province as experimentation area,and the accuracy evaluation of Globeland30 is realized.We calculate the population samples by traditional probability and statistics theory and put the Neyman allocation and distribution according to area ratio combination and assign samples in this way.When the samples are laid,we introduce the landscape shape index on the basis of stratified layout to improve the sample＇s typicality and the reliability of the evaluation results.Using confusion matrix,the overall accuracy of land cover products in Jiangxi province is 81.94%,and the Kappa coefficient is 0.7005.The paper indicates that the quality of the GlobeLand30 in the experimental region is reliable.
出处
《北京测绘》
2017年第01S期271-277,共7页
Beijing Surveying and Mapping
基金
"十三五"国家重点研发计划"典型资源环境要素识别提取与定量遥感技术"(2016YFB0501404)
