Deep learning-based surface crack detection of bridge structures

摘要：

　　Cracks are a potential threat to the safety and endurance of bridges and require abundant inspection work during daily maintenance.Vision-based approaches are newly developed promising methods for crack detection.Traditional edge detection algorithms are easily disturbed by environmental changes and convolutional neural networks(CNN)have achieved a better performance over other artificial neural networks(ANN).This paper aims at applying a fully convolutional network(FCN)based on CNN for the recognition of cracks in images from concrete bridges.Pixel-level labeled image training data are obtained from online databases.Crack images with different surface conditions are tested for the validation of the recognition capacity of the proposed FCN.The recognition results are compared with the edge detection method.The obtained results indicate that the proposed FCN is able to eliminate plenty of noise disturbing the edge detection method with a reliable performance.A promising application of practical use can be expected for the robustness and the ability to display the location and path of cracks in the images.Due to complicated surface conditions and crack forms,more training data with pixel-level labels in real-world situations is required for improvement.

关键词： Structural Health Monitoring Concrete Bridges Crack Detection Fully Convolutional Network

作者: T.Jin X.W.Ye P.Y.Chen

作者单位: Department of Civil Engineering,Zhejiang University,Hangzhou 310058,China

会议类型: 国际会议

会议名称: The 7th World Conference on Structural Control and Monitoring(7WCSCM)(第七届结构控制与监测世界大会)

会议地点: 青岛

会议语种:英文

页码: 1692-1698

在线出版日期: 2018-07-22（万方平台首次上网日期，不代表论文的发表时间）