Estimation of crack width based on shape‐sensitive kernels and semantic segmentation

Lee, Jun S.; Hwang, Sung Ho; Choi, Il; Choi, Yeong-Tae

doi:10.1002/stc.2504

Cited by 52 publications

(27 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The CDN can delineate cracks quickly and accurately using the proposed empirical optimal fusion strategy. Considering the shape characteristics of cracks, Lee et al 39 proposed crack‐like kernel (CK) models to accomplish crack detection and crack width estimation. Their results indicated that the proposed method based on CK model can estimate crack width more accurately and will serve for their designed decision support system 40 for railway maintenance.…”

Section: Introductionmentioning

confidence: 99%

Deep learning‐based classification and instance segmentation of leakage‐area and scaling images of shield tunnel linings

Zhao

Shadabfar

Zhang

et al. 2021

Struct Control Health Monit

View full text Add to dashboard Cite

This paper presents an approach for the integrated process of classification and instance segmentation of leakage-area and scaling images from shield tunnel linings. For this purpose, the previously established dataset of leakage-area images by the authors is enlarged by means of adding scaling ones. Afterwards, data augmentation is implemented to enrich the database in the classification dataset, and the augmented classification dataset contains 5776 images. The instance segmentation dataset is subsequently enlarged through original images without any data augmentation, including 1496 images. Then a residual net with 101 layers (i.e., ResNet-101) is applied to the classification dataset to obtain a model that can identify leakage-area and scaling images from those of shield tunnel linings. The ResNet-101 classification model achieves an accuracy of 93.37% in terms of testing classification dataset. Moreover, a mask region-based convolutional neural network (Mask R-CNN) is utilized to perform instance segmentation of leakage areas and scaling in the images classified by the ResNet-101 model. The segmentation results of the Mask R-CNN model show 96.1% and 95.6% average precision (AP) with intersection over union (IoU) of 0.5 for bounding box and mask predication, respectively. By using the proposed approach, the leakage-area and scaling defects can be automatically classified and quantified with an overall accuracy of 89.3%, which is quite promising compared to the inherent uncertainty in geotechnical engineering. Image mosaicing is finally applied to provide inspectors with the intuitive observation of the location and distribution information of defects on the tunnel lining.

show abstract

Section: Introductionmentioning

confidence: 99%

Deep learning‐based classification and instance segmentation of leakage‐area and scaling images of shield tunnel linings

Zhao

Shadabfar

Zhang

et al. 2021

Struct Control Health Monit

View full text Add to dashboard Cite

show abstract

“…In addition, the geometry and density of the cracks could not be captured accurately using the identified crack patches. To overcome these limitations, recent studies also featured other variations of CNNs that were more robust to image scales and were capable of recognizing cracks in a pixel level, such as region‐based CNN (RCNN), 34,35 mask RCNN, 36,37 and fully convolutional networks (FCN) 38,39 …”

Section: Introductionmentioning

confidence: 99%

Towards automated detection and quantification of concrete cracks using integrated images and lidar data from unmanned aerial vehicles

Yan

Zhong

et al. 2021

Struct Control Health Monit

View full text Add to dashboard Cite

Summary Recent advances in visual sensing technology and unmanned aerial vehicle (UAV) provide an effective tool to capture the as‐is conditions of infrastructure and thus have gained their popularity in infrastructure inspection and documentation. To facilitate this process, several recent studies have proposed automated methods for detecting concrete cracks from UAV‐based images. This study is aimed at proposing a new method for automating both the detection and quantification of concrete cracks. The proposed method advances state‐of‐the‐art image‐based concrete crack detection methods by integrating the RGB images with the lidar data collected by UAV. The key innovations focus on two aspects: (1) by recognizing objects of interest in the lidar data, regions of interest can be extracted automatically from the images; (2) by retrieving the depth information through the lidar data, the actual pixel sizes can be estimated to facilitate both the detection and quantification of concrete cracks. In order to validate the proposed method, a customized UAV platform that was equipped with a high‐resolution camera and a Velodyne VLP‐16 lidar scanner was developed to scan the substructure elements of an in‐service bridge where multiple concrete cracks can be observed. The effectiveness of the proposed method in recognizing and quantifying concrete cracks is validated quantitively against manually annotated images and physical measurements. The results indicate that the proposed approach can recognize crack pixels with an accuracy of 85% on average as well as quantify the recognized concrete cracks with an error less than 10%.

show abstract

“…While the various automated approaches discussed above have many advantages, visual inspection by experienced staff is still a common approach. Inspired by this, researchers have investigated the use of surface image processing to detect surface rail defects [4,25,27,11]. However, this approach has significant challenges in uncontrolled environments due to, e.g., contaminants on the rail surface.…”

Section: Introductionmentioning

confidence: 99%

In-field railhead crack detection using digital image correlation

Meyer¹,

Gren²,

Ahlström³

et al. 2021

Preprint

View full text Add to dashboard Cite

A significant challenge for railway infrastructure managers is to know when and how to maintain rails. Often, the critical missing information is the current health status, in particular information about rail cracks. Therefore, a new method for rail crack detection is proposed. By utilizing a train-mounted camera system, a single measurement train can monitor a large rail network. The system uses Digital Image Correlation (DIC) to measure the strain fields due to rail bending caused by the measurement train. Promising results are obtained under laboratory conditions. The identified cracks are correlated to the actual crack network, characterized by serial-sectioning microscopy. Furthermore, finite element simulations show the method's high sensitivity to crack depths. Knowing the crack depths enable infrastructure managers to optimize the rail maintenance.

show abstract

Estimation of crack width based on shape‐sensitive kernels and semantic segmentation

Cited by 52 publications

References 36 publications

Deep learning‐based classification and instance segmentation of leakage‐area and scaling images of shield tunnel linings

Deep learning‐based classification and instance segmentation of leakage‐area and scaling images of shield tunnel linings

Towards automated detection and quantification of concrete cracks using integrated images and lidar data from unmanned aerial vehicles

In-field railhead crack detection using digital image correlation

Contact Info

Product

Resources

About