Sub-Block Urban Function Recognition with the Integration of Multi-Source Data

Area of Science:

• Urban Studies
• Geographic Information Systems (GIS)
• Remote Sensing

Background:

• Traditional urban functional area (UFA) recognition is limited by singular data sources, incomplete results, and insufficient socioeconomic context.
• Advancements in multi-source big data offer new possibilities for dynamic UFA recognition.

Purpose of the Study:

• To propose a sub-block function recognition framework integrating multi-feature information for UFA classification.
• To enhance the accuracy and comprehensiveness of UFA recognition at the sub-block level.

Main Methods:

• Developed a framework integrating building footprints, point-of-interest (POI) data, and Landsat imagery.
• Employed a random forest model for classifying UFAs at the sub-block level.
• Incorporated three-dimensional (3D) building features and detailed land cover information.

Main Results:

• Achieved significantly higher recognition accuracies for single- and mixed-function areas compared to existing methods in Guangzhou, China.
• Attained an overall accuracy (OA) of 82% for single-function areas, outperforming other models by 8-36%.
• Demonstrated that integrating 3D building and finer land cover features improves UFA recognition accuracy.

Conclusions:

• The proposed method provides a more practical and comprehensive solution for UFA recognition using open-access data.
• The integration of diverse data sources and advanced features is crucial for accurate dynamic UFA mapping.
• This approach supports improved urban structure understanding and planning.