Abstract:Accurately understanding the functions of buildings is crucial for urban monitoring, analysis of urban economic structures, and effectively allocating resources. Previous studies have investigated building function classification using single or dual data sources. However, the complexity of building functions cannot be fully reflected by a limited number of data sources. In addition, the functions of adjacent buildings often exhibit correlations, and contextual information between buildings has been ignored in previous studies. To address these problems, we propose a graph-based neural network (GNN) approach for building function classification that integrates multi-source data and mines contextual information between buildings. This approach initially extracts four types of features related to building functions: morphological features from vector-buildings, visual features from street-view images, spectral features from satellite images, and socio-economic features from points of interest. The buildings are then modeled as a graph, where the nodes and edges represent the buildings and their proximity. Descriptive features of the nodes were obtained by concatenating the aforementioned features. Finally, the constructed graph was fed into the GraphSAmple and aggreGatE (GraphSAGE) model, which is a typical GNN model for building function classification. The experimental results showed that our approach achieved an F1-score of 91.0%, which was 10.3–12.6% higher than that of the three comparison approaches. In addition, ablation experiments using different data sources revealed that the four data sources were complementary to each other and contributed to improving the building function classification. Our strategy provides an alternative and efficient solution for building function classification.

Recognition of building group patterns using GCN and knowledge graph

Building pattern recognition by using an edge-attention multi-head graph convolutional network

Mixed Geometry Message and Trainable Convolutional Attention Network for Knowledge Graph Completion

Inferring High-level Geographical Concepts via Knowledge Graph and Multi-scale Data Integration: A Case Study of C-shaped Building Pattern Recognition

Linear building pattern recognition via spatial knowledge graph

Enhancing building pattern recognition through multi-scale data and knowledge graph: a case study of C-shaped patterns

Mining Location-Aware Discriminative Blocklets for Recognizing Landmark Architectures

Learning to Group Discrete Graphical Patterns

Combining Deep Fully Convolutional Network and Graph Convolutional Neural Network for the Extraction of Buildings from Aerial Images

Optimizing Spatial Relationships in GCN to Improve the Classification Accuracy of Remote Sensing Images

GCN-based multi-scale dual fusion for remote sensing building change detection

Graph isomorphism network with weighted multi‐aggregators for building shape classification

Modelling Irregular Spatial Patterns Using Graph Convolutional Neural Networks

Joint Urban Modeling With Graph Convolutional Networks and Crowdsourced Data: A Novel Approach

An axis-matching approach to combined collinear pattern recognition for urban building groups

Partitioning building groups at multiple scales based on image segmentation

A graph-based neural network approach to integrate multi-source data for urban building function classification

Linear building pattern recognition in topographical maps combining convex polygon decomposition

TriangleConv: A Deep Point Convolutional Network for Recognizing Building Shapes in Map Space

Learning Spatial Interaction Representation with Heterogeneous Graph Convolutional Networks for Urban Land-Use Inference

Learning Graph Convolutional Networks for Multi-Label Recognition and Applications