A UAV-based platform for automated inspection of cracked concrete surfaces

Paper i proceeding, 2019

The manual inspection and maintenance of concrete civil engineering structures is time and resource intensive. In general, these inspections require inspectors to reach areas of the structure that are difficult to access, such as bridge decks and supports. The resulting inspection is difficult to perform, and requires the structure to be closed or otherwise taken out of operation during inspection. Thus, autonomous inspection of these structures in situ is a desirable alternative to current inspection methods. This paper investigates a method for visually detecting damage in concrete structures on a UAV platform. Photogrammetry data acquired from a hexacopter UAV is analyzed by a supervised convolutional neural network (CNN) pre-trained to detect the appearance of cracks in concrete surfaces. In this paper, we scan a reinforced concrete specimen for cracks from a UAV-based tablet employing Dot3D imaging software. Image stability and algorithmic accuracy will be commented upon, as well as future areas for methodological improvement. © International Workshop on Structural Health Monitoring.