UAV remote sensing and deep learning for assessing and optimizing architectural texture in traditional villages

The architectural texture of traditional villages reflects regional culture and is vital for sustainable conservation and renewal. However, for culturally diverse villages, current methods often lack integration of multi-dimensional evaluation and fusion of traditional and modern techniques, limiting data efficiency and analytical rigor. This study proposes a framework combining UAV remote sensing, deep learning (optimized Mask R-CNN), morphological indices, and statistical analysis. Applied to 27 traditional villages in Beijing, results show: (1) the method enables fast, accurate extraction of architectural texture with minimal manual input; (2) villages show heterogeneity in scale, proportion, and orientation, while patterns and boundaries remain stable; (3) texture features are influenced by geography, culture, history, economy, and construction methods; (4) indicators like settlement size and building proportion correlate strongly with other variables, offering insights for spatial planning. This interdisciplinary approach supports scientific evaluation and offers a digital foundation for preserving and optimizing traditional village environments.