SDNET2018 is an annotated image dataset for training, validation, and benchmarking of artificial intelligence based crack detection algorithms for concrete. SDNET2018 contains over 56,000 images of cracked and non-cracked concrete bridge decks, walls, and pavements. The dataset includes cracks as narrow as 0.06 mm and as wide as 25 mm. The dataset also includes images with a variety of obstructions, including shadows, surface roughness, scaling, edges, holes, and background debris. SDNET2018 will be useful for the continued development of concrete crack detection algorithms based on deep convolutional neural networks (DCNNs), which are a subject of continued research in the field of structural health monitoring. The authors present benchmark results for crack detection using SDNET2018 and a crack detection algorithm based on the AlexNet DCNN architecture. SDNET2018 is freely available at https://doi.org/10.15142/T3TD19.

Many state agencies are investigating the use of unmanned aerial systems (UASs) for bridge inspections. Some agencies are receiving pressure from consultants and their own administrations to implement UAS inspections with limited knowledge of their efficacy. This paper studies the feasibility of using UASs for fatigue crack detection in bridges with fracture critical members (FCMs) through real-time and postflight visual inspection. The effects of surface illumination on the maximum crack-to-camera (MCC) distance at which a fatigue crack can be detected were investigated in the laboratory. Mock field inspections evaluated the achievable crack-to-platform (ACP) distance in Global Positioning System (GPS)-denied and windy environments to determine whether known cracks can be identified at achievable standoff distances. Finally, two FCM inspections demonstrated the field performance of UASs in identifying fatigue cracks. Results highlight the importance of camera specifications and surface illumination in determining the required standoff distance of crack detection. Furthermore, the results demonstrate the difficulties in obtaining clear images with unstable UASs in GPS-denied or windy environments. Nevertheless, the best performing platform tested in this study exhibited a performance comparable to an average of 30 human inspectors at a fatigue crack identification training structure. The limited results presented here proved the feasibility of using UASs for fatigue crack detection in FCM inspections of steel bridges and highlighted the shortcomings of UASs for this type of hands-on inspection.

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.

This paper presents a new method to develop stochastic deterioration models using a combination of methods including Markov chains, logistic regression, and classification trees. It is computationally more efficient to use logistic regression with the Markov chain process than it is to use optimization-based approaches, and the former is shown to marginally improve the prediction of condition ratings for small data sets. Annually inspected bridge data are split into groups using a classification tree, and logistic regression is used to determine transition probabilities for a Markov chain process. A case study was conducted to determine the effectiveness of using the proposed logistic regression and Markov chain approach for the small data sets created by the classification tree. Wyoming bridge inspection data were split into 15 subsets based on 5 explanatory variables, and deterioration models were developed for each subset. Error analysis showed that logistic regression performed marginally better than traditional methods when estimating the transition probability matrix when limited data are accessible.

This paper presents a procedure to select explanatory variables, without the influence of human bias, for deterioration model development using publicly available data for bridges in Wyoming. Using this information, including geometric data and climate information, candidate explanatory variables can be converted into normalized numeric values and analyzed prior to the development of deterministic or stochastic deterioration models. The prevailing approach for explanatory variable selection for bridge condition modeling is to use expert opinions solicited from experienced engineers. This may introduce personal human influenced biases into the deterioration modeling process. A framework using least absolute shrinkage and selection operator (LASSO) penalized regression and covariance analysis are combined to compensate for this potential bias. The resultant of LASSO, combined with the cross validation, is used to develop a LASSO solution path that visualizes the relative significance of investigated explanatory variables. In order to analyze yearly inspected data in the framework, the cross-validation analysis is used as a standard for the selection of minimum number of explanatory variables using the solution path. The proposed method is demonstrated through the creation of deterministic deterioration models for deck, superstructure, and substructure for Wyoming bridges and compared to explanatory variables using the expert selection method. The comparison shows a significant decrease in error using the presented framework based on the least squares of relative error.