融合多源数据的城市功能区识别与分析  

Identification and Analysis of Urban Functional Areas by Fusion of Multi-source Data

作　　者：齐广玉 程玮瑜 程朋根[1,3] QI Guangyu;CHENG Weiyu;CHENG Penggen(School of Surveying and Geoinformation Engineering,East China University of Technology,330013,Nanchang,PRC;The First Institute of Architectural Design,Tongji University Architectural Design and Research Institute(Group)Co.Ltd.,200092,Shanghai,PRC;Key Laboratory of Mine Environmental Monitoring and Improving around Poyang Lake,Ministry of Natural Resources,330013,Nanchang,PRC)

机构地区：[1]东华理工大学测绘与空间信息工程学院,南昌330013 [2]同济大学建筑设计研究院(集团)有限公司建筑设计一院,上海200092 [3]自然资源部环鄱阳湖区域矿山环境监测与治理重点实验室,南昌330013

出　　处：《江西科学》2024年第2期289-296,共8页Jiangxi Science

基　　金：国家自然科学基金项目(41861052);江西省自然科学基金面上项目(20202BABL202045)。

摘　　要：城市化的快速发展使城市的空间结构发生变化,合理划分城市功能区有利于监测城市化以及城市规划与管理。遥感影像能反映地物的物理特征,但无法获取其社会经济特征。使用高分二号遥感影像数据、POI数据、夜间灯光数据和建筑物轮廓数据,融合多源数据的特征信息并基于Scikit-Learn机器学习方法实现城市功能区的划分。首先,以道路网为基本研究单元构建交通分析区,将研究区划分为827个地块,然后,结合核密度分析、频数密度法和区域分析,从研究区内提取并融合多源数据的特征信息,基于3种分类模型识别城市功能区。研究结果表明,通过综合利用光谱、纹理构建的BOVW模型、POI和夜间灯光数据构建的社会经济特征和建筑物轮廓数据的景观特征等特征指标,结合随机森林模型的方法,取得了最佳识别结果,其精度高达76.65%。验证了本文方法的可行性和有效性。The rapid development of urbanization has changed the spatial structure of the city,and the rational division of urban functional areas is conducive to the monitoring of urbanization and urban planning and management.Remote sensing images can reflect the physical characteristics of ground objects,but cannot obtain their socioeconomic characteristics.This study uses Gaofen-2 remote sensing image data,POI data,nighttime light data and building outline data to integrate the feature information of multi-source data and implement the division of urban functional areas based on the Scikit-Learn machine learning method.Firstly,the traffic analysis area was constructed with the road network as the basic research unit,and the research area was divided into 827 plots,and then combined with kernel density analysis,frequency density method and regional analysis,extracted and the feature information of multi-source data is integrated to identify urban functional areas based on three classification models.The research results show that the best recognition results are achieved by comprehensively utilizing the BOVW model constructed from spectrum and texture,socioeconomic characteristics constructed from POI and night light data,and landscape characteristics of building outline data,combined with the random forest model method,its accuracy is as high as 76.65%.The feasibility and effectiveness of this method are verified.

关 键 词：高分二号遥感影像 POI 随机森林 城市功能区 

分 类 号：P208[天文地球—地图制图学与地理信息工程]