基于遥感数据集的城市不透水面识别精度比较——以河南省为例

Comparison of Urban Impervious Surface Identification Accuracy Based on Remote Sensing Datasets—A Case Study of Henan Province

摘要: 近年来，众多学者公开发布不同的遥感不透水面数据集产品，在区域和全球范围尺度内，其动态监测能力存在一定程度的差异，因此，对于不同数据集的质量比较与分析显得尤为必要。文章依据2015年Google Earth上的遥感影像数据，通过混淆矩阵来衡量不同数据集的准确性，并使用Kappa系数来评估数据的可靠性。结果表明：2015年GISA数据集在河南省的总体分类精度达到93.7%，Kappa系数为0.874。GISA数据集在数据质量上，优于GAIA和GISD数据集，更适合河南省区域内不透水面的研究。文中通过标准差椭圆分析方法研究1985~2019年间河南省不透水面的时空演变趋势，发现河南省不透水面发展在时空分布上呈现扩张趋势；城市不透水面不断扩张，主要由城市中心区域向四周扩展。研究结果可为河南省区域不透水面数据集的选取及分析提供一定的数据参考。

Abstract: In recent years, many scholars have publicly released different remote sensing impervious surface dataset products. At the regional and global scales, there are some differences in their dynamic monitoring capabilities. Therefore, it is particularly necessary to compare and analyze the quality of different datasets. Based on the remote sensing image data on Google Earth in 2015, this paper measures the accuracy of different datasets by confusion matrix, and uses Kappa coefficient to evaluate the reliability of the data. The results show that the overall classification accuracy of the GISA dataset in Henan Province in 2015 reached 93.7%, and the Kappa coefficient was 0.874. GISA dataset is superior to GAIA and GISA datasets in data quality, and is more suitable for the study of impervious surface in Henan Province. In this paper, the spatial and temporal evolution trend of impervious surface in Henan Province from 1985 to 2019 is studied by standard deviation ellipse analysis method, and it is found that the development of impervious surface in Henan Province shows an expansion trend in spatial and temporal distribution. Urban impervious surface continues to expand, mainly from the city center to the surrounding area. The research results can provide some data reference for the selection and analysis of regional impervious surface datasets in Henan Province.