Detection of exposed steel rebars based on deep-learning techniques and unmanned aerial vehicles

• Published in: Automation in construction 2022-07, Vol.139, p.104324, Article 104324
• Main Authors: Santos, R., Ribeiro, D., Lopes, P., Cabral, R., Calçada, R.
• Format: Article
• Language: English
• Subjects: Artificial neural networks; Concrete structures; Convolutional neural network (CNN); Deep learning; Exposed rebar; Flaw detection; Inspection; Rebar; Reinforced concrete (RC); Reinforcing steels; Remote inspection; Three dimensional models; Unmanned aerial vehicles; Unmanned aerial vehicles (UAVs)
• ISSN: 0926-5805; 1872-7891
• DOI: 10.1016/j.autcon.2022.104324

In recent years deep-learning techniques have been developed and applied to inspect cracks in RC structures. The accuracy of these techniques leads to believe that they may also be applied to the identification of other pathologies. This article proposes a technique for automated detection of exposed steel rebars. The tools developed rely on convolutional neural networks (CNNs) based on transfer-learning using AlexNet. Experiments were conducted in large-scale structures to assess the efficiency of the method. To circumvent limitations on the proximity access to structures as large as the ones used in the experiments, as well as increase cost efficiency, the image capture was performed using an unmanned aerial system (UAS). The final goal of the proposed methodology is to generate orthomosaic maps of the pathologies or structure 3D models with superimposed pathologies. The results obtained are promising, confirming the high adaptability of CNN based methodologies for structural inspection. •Methodology for automatic detection of exposed steel rebars in RC structures.•Support of Unmanned Aerial Vehicles and Artificial Intelligence.•Region Convolutional Neural Network (R-CNN) based on a dedicated image database.•Orthoimage mosaics with georeferenced identification of the surface anomalies.•Application on large-scale structures: industrial building and telecommunications tower.