Debonding Defect Recognition of Building Decoration Layers by UAV Thermography

The phenomenon of building decorative layers (BDLs) falling off of exterior walls is quite common, and is of great concerns to human safety. In this study, a rotor unmanned aerial vehicle (UAV) equipped with an infrared thermal camera is used as the working platform to detect debonding BDL defects to obtain the change in law of its thermography imagery. Based on the analysis of the temperature field and shape characteristics of thermography images of BDLs, an image segmentation method for debonding defects based on fuzzy clustering is proposed, and a shape feature vector set of debonding BDL defects is constructed. Therefore, a feature learning model for debonding defect recognition and a calculation method for the actual area of debonding defects based on support vector machines are established. Finally, a case study of the teaching building inspection with several peeling veneers is carried out to demonstrate the effectiveness of the proposed method. Compared with the manual test, the results show that the small-sample machine-learning algorithm for debonding defect recognition based on prior feature law has advantages, and can effectively reduce the occurrence of accidents presenting potential practical applications.