A Robot System for Rapid and Intelligent Bridge Damage Inspection Based on Deep-Learning Algorithms

Abstract

Large numbers of bridges have already suffered various types of damage but still operate all year round without proper treatment. Conducted primarily manually, the routine bridge inspections are ineffective in detecting potential damage in time due to a lack of relevant instruments and equipment, particularly modern measures. In this study, a rapid and intelligent bridge inspection system that integrates multiple modules and deep learning algorithms was established. First, the robot inspection equipment is established. Then, the You Only Look Once version 3 (YOLOv3) object detection algorithm is employed to classify four types of defects from the acquired data. Finally, an image segmentation algorithm is used to identify crack defects at a pixel level. Experimental results reveal that the proposed system can be effectively applied to accurately locate defects (e.g., cracks, spalls, exposed tendons, and free lime) and identify cracks at a pixel level on various types of bridges without affecting traffic.