城市地质2025，Vol.20Issue(3)：336-343,8.DOI:10.3969/j.issn.2097-3764.2025.03.007
基于深度学习的北京市中心城区土地分类及降水入渗系数研究
Study on Land use classification and precipitation infiltration coefficient in central urban Beijing based on deep learning
摘要
Abstract
This study employs high-resolution remote sensing imagery and a deep learning-based semantic segmentation model to classify urban surfaces(impervious,green space,water bodies)in Beijing's central urban area.Empirically derived infiltration coefficients were assigned to each land cover type to generate a spatially explicit precipitation infiltration coefficient map.Results reveal:1.Low overall infiltration capacity,with coefficients predominantly ranging from 0.1 to 0.25.2.Distinct spatial heterogeneity,exhibiting a"higher periphery,lower center"pattern.Our approach demonstrates that deep learning-enhanced remote sensing classification significantly improves accuracy in complex urban land cover identification and provides high-quality input data for robust precipitation infiltration estimation.The methodology shows strong potential for dynamic monitoring of urban land use and infiltration processes,facilitating timely data updates.The resulting infiltration coefficient map effectively quantifies the spatial variability of surface infiltration under urbanization,offering scientific evidence and technical support for groundwater management and ecological spatial planning.关键词
深度学习/遥感/土地分类/降水入渗系数/北京市中心城区/地下水补给Key words
deep learning/remote sensing/land classification/rainfall infiltration coefficient/central urban Beijing/groundwater recharge引用本文复制引用
胡仲毅..基于深度学习的北京市中心城区土地分类及降水入渗系数研究[J].城市地质,2025,20(3):336-343,8.基金项目
北京市地下水监测网运行项目(11000022T000000440194)资助 （11000022T000000440194）
