Comparison of deep convolutional neural networks and edge detectors for image-based crack detection in concrete

Dorafshan, Sattar; Thomas, Robert J.; Maguire, Marc

doi:10.1016/j.conbuildmat.2018.08.011

Cited by 504 publications

(276 citation statements)

References 48 publications

(56 reference statements)

Supporting

Mentioning

235

Contrasting

Unclassified

Order By: Relevance

“…A robotic crack inspection and mapping (ROCIM) system is proposed to provide an overall solution to the bridge deck crack inspection. Edge detection algorithms such as Roberts, Prewitt, Sobel, Laplacian of Gaussian, Butterworth, and Gaussian were used to detect cracks in 3420 concrete sub-images [9]. These methods can accurately detect 53-79% of cracked pixels, but lead to residual noise in the final binary image.…”

Section: Related Workmentioning

confidence: 99%

Automatic Tunnel Crack Detection Based on U-Net and a Convolutional Neural Network with Alternately Updated Clique

et al. 2020

Sensors

View full text Add to dashboard Cite

Regular crack inspection of tunnels is essential to guarantee their safe operation. At present, the manual detection method is time-consuming, subjective and even dangerous, while the automatic detection method is relatively inaccurate. Detecting tunnel cracks is a challenging task since cracks are tiny, and there are many noise patterns in the tunnel images. This study proposes a deep learning algorithm based on U-Net and a convolutional neural network with alternately updated clique (CliqueNet), called U-CliqueNet, to separate cracks from background in the tunnel images. A consumer-grade DSC-WX700 camera (SONY, Wuxi, China) was used to collect 200 original images, then cracks are manually marked and divided into sub-images with a resolution of 496 × 496 pixels. A total of 60,000 sub-images were obtained in the dataset of tunnel cracks, among which 50,000 were used for training and 10,000 were used for testing. The proposed framework conducted training and testing on this dataset, the mean pixel accuracy (MPA), mean intersection over union (MIoU), precision and F1-score are 92.25%, 86.96%, 86.32% and 83.40%, respectively. We compared the U-CliqueNet with fully convolutional networks (FCN), U-net, Encoder–decoder network (SegNet) and the multi-scale fusion crack detection (MFCD) algorithm using hypothesis testing, and it’s proved that the MIoU predicted by U-CliqueNet was significantly higher than that of the other four algorithms. The area, length and mean width of cracks can be calculated, and the relative error between the detected mean crack width and the actual mean crack width ranges from −11.20% to 18.57%. The results show that this framework can be used for fast and accurate crack semantic segmentation of tunnel images.

show abstract

Section: Related Workmentioning

confidence: 99%

Automatic Tunnel Crack Detection Based on U-Net and a Convolutional Neural Network with Alternately Updated Clique

et al. 2020

Sensors

View full text Add to dashboard Cite

show abstract

“…Image processing algorithms can improve the accuracy and efficiency of autonomous inspections by either (a) enhancing images to improve ease of human detection of defects or (b) autonomously identifying defects. Additionally, edge detectors are used in combination with more contemporary techniques such as deep learning convolutional neural networks for UAS applications [27], reducing false positive cases by 20 times compared to sole use of edge detectors [28].…”

Section: Introductionmentioning

confidence: 99%

“…Crack detection algorithms can emphasize edges by applying filters in either the spatial or frequency domain. Even though use of edge detectors for crack detection goes back to the early 2000s [29], these methods have been used in the past and are still being used in recent studies because of their simplicity and pixel-based detection of cracks [15,28,[30][31][32][33][34][35][36][37][38][39][40][41][42]. Even in emerging applications of supervised machine learning methods, edge detectors are still considered in practice since they do not require expensive annotated training datasets.…”

Section: Introductionmentioning

confidence: 99%

Benchmarking Image Processing Algorithms for Unmanned Aerial System-Assisted Crack Detection in Concrete Structures

2019

View full text Add to dashboard Cite

This paper summarizes the results of traditional image processing algorithms for detection of defects in concrete using images taken by Unmanned Aerial Systems (UASs). Such algorithms are useful for improving the accuracy of crack detection during autonomous inspection of bridges and other structures, and they have yet to be compared and evaluated on a dataset of concrete images taken by UAS. The authors created a generic image processing algorithm for crack detection, which included the major steps of filter design, edge detection, image enhancement, and segmentation, designed to uniformly compare different edge detectors. Edge detection was carried out by six filters in the spatial (Roberts, Prewitt, Sobel, and Laplacian of Gaussian) and frequency (Butterworth and Gaussian) domains. These algorithms were applied to fifty images each of defected and sound concrete. Performances of the six filters were compared in terms of accuracy, precision, minimum detectable crack width, computational time, and noise-to-signal ratio. In general, frequency domain techniques were slower than spatial domain methods because of the computational intensity of the Fourier and inverse Fourier transformations used to move between spatial and frequency domains. Frequency domain methods also produced noisier images than spatial domain methods. Crack detection in the spatial domain using the Laplacian of Gaussian filter proved to be the fastest, most accurate, and most precise method, and it resulted in the finest detectable crack width. The Laplacian of Gaussian filter in spatial domain is recommended for future applications of real-time crack detection using UAS.

show abstract

“…In order to know the current development of edge detection, especially these seven algorithms, a search based on Web of Science data is shown in Table 1 with searching "the names of these algorithms AND 'edge detection'" with "Topic" selection from 2009 to 2018. As this paper is related to civil engineering, the number related to it is listed [32][33][34][35][36]. Table 1 reflects that Sobel and Canny are much more discussed than the other five.…”

Section: Introductionmentioning

confidence: 99%

Evaluation of Classical Operators and Fuzzy Logic Algorithms for Edge Detection of Panels at Exterior Cladding of Buildings

et al. 2019

View full text Add to dashboard Cite

The automated process of construction defect detection using non-contact methods provides vital information for quality control and updating building information modelling. The external cladding in modular construction should be regularly controlled in terms of the quality of panels and proper installation because its appearance is very important for clients. However, there are limited computational methods for examining the installation issues of external cladding remotely in an automated manner. These issues could be the incorrect sitting of a panel, unequal joints in an elevation, scratches or cracks on the face of a panel or dimensions of different elements of external cladding. This paper aims to present seven algorithms to detect panel edges and statistically compare their performance through application on two scenarios of buildings in construction sites. Two different scenarios are selected, where the building façades are available to the public, and a sample of 100 images is taken using a state-of-the-art 3D camera for edge detection analysis. The experimentation results are validated by using a series of computational error and accuracy analyses and statistical methods including Mean Square Error, Peak Signal to Noise Ratio and Structural Similarity Index. The performance of an image processing algorithm depends on the quality of images and the algorithm utilised. The results show better performance of the fuzzy logic algorithm because it detects clear edges for installed panels. The applications of classical operators including Sobel, Canny, LoG, Prewitt and Roberts algorithms give similar results and show similarities in terms of the average of errors and accuracy. In addition, the results show that the minor difference of the average of the error and accuracy indices for Sobel, Canny, LoG, Prewitt and Roberts methods between both scenarios are not statistically significant, while the difference in the average of the error and accuracy indices for RGB-Sobel and Fuzzy methods between both scenarios are statistically significant. The accuracy of the algorithms can be improved by removing unwanted items such as vegetation and clouds in the sky. The evaluated algorithms assist practitioners to analyse their images collected day to day from construction sites, and to update building information modelling and the project digital drawings. Future work may need to focus on the combination of the evaluated algorithms using new data sets including colour edge detection for automatic defect identification using RGB and 360-degree images.

show abstract

Comparison of deep convolutional neural networks and edge detectors for image-based crack detection in concrete

Cited by 504 publications

References 48 publications

Automatic Tunnel Crack Detection Based on U-Net and a Convolutional Neural Network with Alternately Updated Clique

Automatic Tunnel Crack Detection Based on U-Net and a Convolutional Neural Network with Alternately Updated Clique

Benchmarking Image Processing Algorithms for Unmanned Aerial System-Assisted Crack Detection in Concrete Structures

Evaluation of Classical Operators and Fuzzy Logic Algorithms for Edge Detection of Panels at Exterior Cladding of Buildings

Contact Info

Product

Resources

About