Fast Detection of Missing Thin Propagating Cracks during Deep-Learning-Based Concrete Crack/Non-Crack Classification

Abstract

Existing deep learning (DL) models can detect wider or thicker segments of cracks that occupy multiple pixels in the width direction, but fail to distinguish the thin tail shallow segment or propagating crack occupying fewer pixels. Therefore, in this study, we proposed a scheme for tracking missing thin/propagating crack segments during DL-based crack identification on concrete surfaces in a computationally efficient manner. The proposed scheme employs image processing as a preprocessor and a postprocessor for a 1D DL model. Image-processing-assisted DL as a precursor to DL eliminates labor-intensive labeling and the plane structural background without any distinguishable features during DL training and testing; the model identifies potential crack candidate regions. Iterative differential sliding-window-based local image processing as a postprocessor to DL tracks missing thin cracks on segments classified as cracks. The capability of the proposed method is demonstrated on low-resolution images with cracks of single-pixel width, captured using unmanned aerial vehicles on concrete structures with different surface textures, different scenes with complicated disturbances, and optical variability. Due to the multi-threshold-based image processing, the overall approach is invariant to the choice of initial sensitivity parameters, hyperparameters, and the sequence of neuron arrangement. Further, this technique is a computationally efficient alternative to semantic segmentation that results in pixelated mapping/classification of thin crack regimes, which requires labor-intensive and skilled labeling.

Inspection Robot and Wall Surface Detection Method for Coal Mine Wind Shaft

“…Secondly, the systematic research on the robot-assisted visual inspection method of the shaft lining condition is completed. Considering that the visual inspection of the shaft lining condition is not specific enough, and the detection of crack defects focuses on concrete, roads [24] and bridges [14], as mentioned in [20,21,25], the nondestructive inspection technology based on vision is rarely used to detect shaft lining cracks. Therefore, by using the characteristics of [13,20,21,24] for reference, a CNN model for the defect recognition of shaft wall images is constructed, and the accurate classification of shaft wall images (health, cracks and spalling) is realized, with a recognition accuracy as high as 97.1%.…”

“…Zhang [23] and others put forward a six-layer CNN to detect and characterize road cracks, which can improve the accuracy of road crack detection. The scheme proposed by Kolappan Geetha et al [24] adopted image processing as the preprocessing and post-processing of a 1D DL model. As the predecessor of DL, image-processing-aided DL eliminates labor-intensive marks and a plane structure background without any distinguishing features during DL training and testing.…”

Deep Learning for Structural Health Monitoring: Data, Algorithms, Applications, Challenges, and Trends

“…However, when processing vibration signals using 2D-CNN, it is necessary to convert one-dimensional data into two-dimensional data. On this basis, One-Dimensional Convolutional Neural Networks (1D-CNN) with simple architecture and low computational complexity are applied to SHM to directly process 1D data for crack detection [56,57], corrosion detection [58], multi-type damage identification [1], abnormal data detection [59,60], etc. In addition to CNN, Recurrent Neural Networks (RNN) can identify the time features of data.…”

Convolutional Neural Network for Predicting Failure Type in Concrete Cylinders During Compression Testing

“…Therefore, the speed and accuracy of classification algorithms constitute key elements for the detection of damage in concrete. Several methods use deep learning models that automate the process [5,12], using inspection techniques based on computer vision for the classification of crack types [7].…”