Rebar detection and localization for bridge deck inspection and evaluation using deep residual networks
Rebar detection and localization for bridge deck inspection and evaluation using deep residual networks
复制标题
DOI：
10.1016/j.autcon.2020.103393
复制
发表时间：
2020-12-01
影响因子：
10.3
通讯作者：
Tran, Khiem
中科院分区：
文献类型：
--
作者：
Ahmed, Habib;La, Hung Manh;Tran, Khiem
民用基础设施的结构健康监测(SHM)和无损评估(NDE)在过去几十年中一直是一个活跃的研究领域。由于成本上升、安全问题和人工检测方法的误差，人们提出了自动化的桥梁检测和维护方法。本研究的目的是利用有监督(深度残差网络)和无监督(KMeans聚类)技术开发一种钢筋自动检测和定位系统。使用探地雷达(GPR)传感器从九座桥梁收集了数据。所提出的钢筋检测和定位系统的性能已经在广泛的性能指标上进行了评估，这强调了所提出的技术优于现有方法。结果表明，网络层数、训练时间与其他性能指标呈正相关。该系统的整体性能也依赖于数据集，受噪声伪影、反射和钢筋轮廓视觉质量等因素的影响。
Structural Health Monitoring (SHM) and Nondestructive Evaluation (NDE) of civil infrastructure has been an active area of research for the past few decades. Due to rising costs, safety issues and error of human inspection methods, automated methods for bridge inspection and maintenance are being proposed. The purpose of this research is to develop an automated rebar detection and localization system utilizing supervised (Deep Residual Networks) and unsupervised (Kmeans clustering) techniques. Data has been collected from nine bridges using Ground Penetrating Radar (GPR) sensors. The performance of the proposed rebar detection and localization system has been evaluated on a wide-range of performance metrics, which emphasize the superior performance of the proposed technique over existing methods. The results reveal positive correlation between number of layers of networks, training time and other performance metrics. The overall performance of the proposed system is also dataset-dependent with factors such as noise artefacts, reflections and visual quality of rebar profiles.