Drone-Based Façade Defect Detection via Decision-Level Fusion of Thermal Infrared and Visible Images

Urban façade deterioration, especially tile delamination and related surface damage, poses significant safety risks in dense urban environments. Conventional inspections based on manual visual assessment and acoustic sounding are labor-intensive, hazardous, and subjective. This paper presents an autonomous UAV-based façade inspection framework that combines oblique-photogrammetry-driven path planning with decision-level fusion of infrared and visible imagery for defect identification. A 3D model of the target building is reconstructed to extract geometric features for generating safe flight trajectories at a fixed standoff distance of 10 m, supporting near-complete and spatially consistent dual-modal data acquisition under the planned inspection route. Modality-specific defects are then segmented from RGB and thermal data using deep-learning models, and the fused defect map is produced through decision-level fusion of surface appearance cues and subsurface thermal signatures. A field deployment on a building demonstrates the feasibility and efficiency of the framework, completing the inspection of approximately 4,200 m² of façade area using 208 waypoints within 50 minutes of total on-site operation. The results show that the proposed method provides a practical and safer solution for façade condition assessment and maintenance prioritization.