Intelligent Recognition of Multiple Diseases in Steel Bridges Based on Improved Mask R-CNN

Steel bridges are integral to modern transportation infrastructure, but they may have structural issues such as rusting, falling bolts, and other deformations due to long-term operation and environmental factors. Traditional methods for identifying these defects are time-consuming and subjective, and they require manual intervention. To address this issue, this paper introduces an intelligent steel-bridge disease detection method utilizing deep learning. High-resolution images of affected areas on steel bridges are collected using unmanned aerial vehicles (UAVs), and these images are subsequently enhanced and annotated to establish a steel bridge disease image library for training and testing. The Mask R-CNN algorithm is employed to build a detection model, in which the backbone network is modified from ResNet101 to VoVNet for performance enhancement. The optimized model exhibits an average precision of 0. 84 and 0. 59 for intersection over union values of 0. 5 and 0. 5 : 0. 95. respectively, demonstrating a 10% improvement over the pre-optimization model. The application of the model to Shangxin Bridge showed detection accuracies of 89. 3%, 85. 7% . and 73. 1 % for coating corrosion, bolt corrosion, and bolt detachment, respectively. The results highlight the efficacy of the refined model in steel bridge disease identification. The combination of UAV and deep learning has significant scientific and engineering value. © 2024 Chang'an University. All rights reserved.