Efficient comparative analysis of traditional buildings in traditional villages using UAV

With rapid urbanization, traditional villages and their buildings face increasing risks of disappearance, requiring more efficient and scientific conservation approaches. This study focuses on 10 representative traditional villages in Guangzhou, China, and develops an integrated digital analysis workflow combining unmanned aerial vehicle (UAV)-based low-altitude remote sensing, 3D modeling, image segmentation, and machine learning for large-scale and accurate identification of traditional buildings. Using centimeter-level 3D models and high-resolution orthophotos, the spatial layout and architectural details of the villages were reconstructed to support digital heritage management. A total of 3414 traditional buildings were identified and 30 key features extracted, enabling comparative analysis of spatial distribution, building scale, roofing materials, roof slope, and plan morphology. The results reveal diverse village characteristics shaped by geography, historical evolution, and functional zoning, showing variations in spatial compactness, building size, and roof form. By establishing a multidimensional regional dataset, this study enhances the efficiency and scientific basis of heritage analysis. The integration of image segmentation and machine learning provides an automated, reproducible framework for large-scale comparative studies, offering theoretical and data support for the digital conservation and sustainable revitalization of Lingnan traditional village heritage.