Concrete crack detection with handwriting script interferences using faster region‐based convolutional neural network

Abstract

The current bridge maintenance practice generally involves manual visual inspection, which is highly subjective and unreliable. A technique that can automatically detect defects, for example, surface cracks, is essential so that early warnings can be triggered to prevent disaster due to structural failure. In this study, to permit automatic identification of concrete cracks, an ad‐hoc faster region‐based convolutional neural network (faster R‐CNN) was applied to contaminated real‐world images taken from concrete bridges with complex backgrounds, including handwriting. A dataset of 5,009 cropped images was generated and labeled for two different objects, cracks and handwriting. The proposed network was then trained and tested using the generated image dataset. Four full‐scale images that contained complex disturbance information were used to assess the performance of the trained network. The results of this study demonstrate that faster R‐CNN can automatically locate crack from raw images, even with the presence of handwriting scripts. For comparative study, the proposed network is also compared with You Only Look Once v2 detection technique.

Uncertainty‐assisted deep vision structural health monitoring

“…Pavement defects and road conditions have been studied using different algorithms including probability generative models and support vector machines (Ai, Jiang, Kei, & Li, 2018), convolutional neural networks (CNNs) (Bang, Park, Kim, & Kim, 2019), and recurrent neural networks (A. Zhang et al., 2017). Identifying cracks have also been the focus of several SHM studies using either bounding boxes (Cha, Choi, & Büyüköztürk, 2017; Deng, Lu, & Lee 2020; Xue & Li, 2018) or semantic segmentation (Sajedi & Liang, 2019a; Yang et al., 2018). Other types of structural defects, such as delamination (Cha, Choi, Suh, Mahmoudkhani, & Büyüköztürk, 2018), cavity (R. Li, Yuan, Zhang, & Yuan, 2018; C. Zhang, Chang, & Jamshidi , 2019), fatigue cracks (Hoskere, Narazaki, Hoang, & Spencer, 2018), and efflorescence (S. Li, Zhao, & Zhou, 2019), or a subset of them are identified using deep learning architectures.…”

Automated crack evaluation of a high‐rise bridge pier using a ring‐type climbing robot

“…Zhang et al, 2017) and concrete structures Gibb, La, & Louis, 2018;Jang, Kim, & An, 2019;Ni, Zhang, & Chen, 2019). Then, region-CNNs have been developed to localize cracks using the vision images collected from pavements (Maeda, Sekimoto, Seto, Kashiyama, & Omata, 2018), tunnels (Xue & Li., 2018), bridges (Deng, Lu, & Lee, 2019), and bridges and buildings (Cha, Choi, Suh, Mahmoudkhani, & Büyüköztürk, 2017). More recently, a semantic segmentation technique has been proposed to automatically classify the crack regions of road pavements (Bang, Park, Kim, & Kim, 2019) and concrete structures (Dung & Anh, 2019;S.…”

Crack Detection from a Concrete Surface Image Based on Semantic Segmentation Using Deep Learning

“…There are also the studies by Minami et al (2019aMinami et al ( , 2019b, in which they improved the performance of photographic devices and performed the CNN. Other studies on crack detection using CNNs have also been conducted by, for example, Cha et al (2017Cha et al ( , 2018, Silva and Lucena (2018), Xue and Li (2018), Deng et al (2019), Jiang and Zhang (2019), Li and Zhao (2019) and Zhang et al (2019). Although several crack detection techniques using deep learning have been proposed so far as above, practical development is still in the early stage, since the technology itself is still immature.…”