WCSE

Intelligent Inspection Robotic System for Bridge Cable Defect Identification and Positioning

Abstract— With a large number of cable-stayed bridges being built and applied, these cables need regular inspection and maintenance to ensure traffic safety. Cable inspection robots can effectively replace traditional manual detection, and the rapid identification of defects is very important to improve inspection efficiency. This paper proposes an intelligent robotic system for bridge cable defect identification and positioning. The 360-degree cable surface images are captured simultaneously by four cameras installed around the developed inspection robot. Based on the Mask R-CNN and image processing, the robotic system can quickly identify the surface defects of cables. The deep learning network is mainly composed of multiple layers of convolutional neural networks, which can achieve classification, regression and pixel-level mask generation of cable surface defect objects. The location information of these identified cable defects will be measured and recorded by ultra-wide band (UWB) positioning components. These positioning data are sent to Kalman filter for optimal estimation. Experimental results indicate that the inspection robotic system can successfully and quickly identify cable defects, and the average processing speed of each cable image is about 0.16s. The robot can realize high-precision defect positioning, and the positioning error is within ±2cm. The application of the robotic system is conducive to confirming the safety of bridges, improving the efficiency of cable inspection, and providing support for the subsequent maintenance.

Index Terms— bridge maintenance, deep learning, inspection robot, defect identification.

Jie Li
School of Mechanical Engineering, Southeast University, China; College of Automation, Nanjing University of Posts and Telecommunications, China
Beibei Li
NO.703 Research Institute of CSSC, China
Chunlei Tu
School of Mechanical Engineering, Southeast University, China
Mengqian Tian
School of Mechanical Engineering, Southeast University, China
Xingsong Wang
School of Mechanical Engineering, Southeast University, China

[Download]

Cite: Jie Li, Chunlei Tu, Xingsong Wang, Mengqian Tian, " Intelligent Inspection Robotic System for Bridge Cable Defect Identification and Positioning, " WCSE 2022 Spring Event: 2022 9th International Conference on Industrial Engineering and Applications, pp. 427-436, Sanya, China, April 15-18, 2022.