Accurately mapping social functional zones of urban green spaces by integrating remote sensing images and crowd-sourced geospatial data

Abstract

Both the physical features and social functions of urban green spaces (UGSs) are crucially important to the ecological and social benefits of urban residents. Increasing attention has been focused on exploring how UGS social functions affect the ecological and social benefits of urban residents, but the social functional classification of UGSs has rarely been studied, and related efficient classification methods are urgently needed. Thus, a novel methodological framework for accurately mapping UGS social functional zones was proposed by integrating remote sensing images, crowd-sourced geospatial data (i.e. point of interest data, the OpenStreetMap road network, and Baidu Map boundary), and a deep learning algorithm. A sequence of combination experiments and ablation experiments were designed for performance validation and for quantifying the contributions of individual crowd-sourced geospatial data to UGS social functional classification. The results showed that the proposed methodological framework can precisely and effectively map UGS social functional zones and that all kinds of crowd-sourced geospatial data contributed to improving the accuracy of UGS social functional classification. This study can assist planners and government departments in the rapid monitoring, effective management, and scientific planning of UGS social functional zones by providing accurate data sources and an effective mapping tool.

Keywords

• Urban green space
• social functional zone
• point of interest
• openstreetmap
• remote sensing