Deep learning-driven pathology detection and analysis in historic masonry buildings of Suzhou

Suzhou’s modern masonry buildings hold substantial historical significance, yet they face escalating risks of deterioration due to regional climate fluctuations and anthropogenic influences. Prompt detection of these issues is essential for effective conservation and restoration. This study integrates UAV technology and deep learning for pathology detection, focusing on five categories: material loss (ML), discoloration and deposits (DD), cracks (CR), surface spalling (SS), and biological invasion (BI). The method integrated 3D scanning and oblique photogrammetry to enable automated facade analysis, effectively addressing the limitations of manual inspection. A case study on Soochow Hospital demonstrated its effectiveness, using over 1200 facade images, with 781 for detection. The model achieved mAP@50 scores of 78% (ML), 47.1% (DD), 48.3% (CR), and 52.2% (SS), meeting conservation needs. This approach ultimately provides valuable technical support for the preservation of Suzhou’s masonry buildings and offers new insights into the conservation of modern masonry heritage.