Using Remote Sensing Data and Graph Theory to Identify Polycentric Urban Structure

Abstract

Polycentric urban structures determine the combination and correlation of urban resources. In the past, nighttime light data were often used to identify the center locations, but the borders of polycentric urban regions (PURs) could not be obtained. Using multisource remote sensing data and graph, this research proposes an effective method for polycentric structure identification. First, we regard nighttime light data as a continuous mathematical surface, which can be constructed as nighttime light intensity graphs (NLIGs). Then, the space-optimized Girvan–Newman (SGN) method is proposed to detect the communities, and the eigenvector centrality (EC) and gray value are used to discover the central node of each community. Finally, the geographical location mapping (GLM) between Landsat 8 data segmentation objects and nodes is established, and the PURs and centers can be mapped to the communities and central nodes. This study took Shenyang, Chengdu, and Xi’an as study areas and used monthly Visible Infrared Imaging Radiometer-National Polar-orbiting Partnership (NPP-VIIRS) data in April 2019 and Landsat 8 data in January and August 2019. The average accuracies of PURs and centers identified by the proposed method were 86.24% and 72.5%, respectively. The developed method can provide technical support and data support for urban planning.