A Method for Detecting Pathologies in Concrete Structures Using Deep Neural Networks

Pathologies in concrete structures, such as cracks, splintering, efflorescence, corrosion spots, and exposed steel bars, can be visually evidenced on the surface. This paper proposes a method for automatically detecting these pathologies from images of structure. The proposed uses deep neural networks to detect images. results time savings error reduction. presents wide-angle containing overall structure also specific pathology identification task cropped region pathology. Identifying images, classification could performed with 99.4% accuracy using cross-validation classifying cracks. Wide no, one, or several same image, case detection, analyzed YOLO network identify five classes. detection were measured mAP, mean Average Precision, classes pathology, reaching 11.80% fissure, 19.22% fragmentation, 5.62% 27.24% bar, 24.44% corrosion. Pathology photos optimized learning.