A GCNN ‐Based Method for Functional Zone Recognition by Integrating Building Spatial Morphology and Courtyard‐Level Context
Zheng Wu, Zhaoxin Dai, Jiping Liu, Chuangqi Wu, Xin Ma, Lining Zhu, Chengcheng Zhang · Transactions in GIS · 2025
ABSTRACT The spatial distribution of functional areas plays a crucial role in urban planning, which is essential for optimizing the spatial configuration of urban land. Urban functional zones involve numerous socioeconomic factors and intricate environments, along with a diverse distribution of internal buildings; conventional convolutional neural networks cannot accurately represent and capture the complex spatial proximity and local structure within these regions. In this study, a novel approach for classifying urban functional zones is proposed using a graph convolutional neural network method. This approach innovatively integrates multidimensional information, including building geometric structures, human activity, and POI heat, by introducing a graph structure to describe the attributes of buildings within the functional zone and their adjacency relationships. This approach effectively addresses the challenges associated with automatic feature extraction and enhances recognition accuracy in functional zone analysis. The experimental findings indicate that the proposed approach achieves excellent results in classifying functional zones, demonstrating substantial improvements compared to traditional rule‐based statistical approaches and machine learning methods. This approach has the potential to solve the problem that it is difficult to accurately model due to the complex spatial structure of urban functional zones.