UAV based falling-object risk assessment of damaged building façades using visible and thermal infrared images

Extreme wind, heavy rainfall, solar radiation, and thermal cycling can accelerate the degradation of interfacial bonding in existing building façade systems. Conventional manual hammer sounding, close-range gondola inspections, and image surveys based on a single modality are often inefficient, hazardous, and difficult to quantify spatially, limiting their applicability to rapid screening of large façade areas and emergency assessment around extreme weather events. This study therefore proposes a falling-object risk assessment method for damaged building façades using UAV visible light and thermal infrared images. The method is guided by the damage evolution from hollowing and spalling to falling object risk. A dual-branch recognition strategy is adopted, in which visible light images are used to detect apparent spalling defects, while thermal infrared images are used to identify thermal anomalies associated with hollowing. The detected defect masks are then mapped onto a photogrammetric model to generate spatial defect objects with true surface area, elevation, façade location, and spatial relationship attributes. Based on the mapped defect area, potential falling impact energy, composite risk regions, exposure conditions, and extreme weather triggers, a screening framework that couples area and energy is established. A case study of a building with tile cladding shows that the proposed method can transform UAV inspection outputs from image-level defect masks into a spatial falling-object risk inventory. The results demonstrate its applicability to quantitative defect assessment, risk prioritization, and façade maintenance.