Innovative Method for Detecting Underwater Concrete Cracks

By Nidhi DhullReviewed by Susha Cheriyedath, M.Sc.Nov 4 2024

A recent article published in the Alexandria Engineering Journal presents a new, non-contact way to detect underwater concrete cracks using a combination of monocular vision and fractal science. This method, which uses computer vision and enhanced image processing, establishes a practical framework for measuring cracks underwater and assessing the overall safety of these concrete structures.

The Problem: Cracks in Underwater Concrete

Concrete is widely used in underwater engineering. However, cracks in underwater concrete structures can disrupt stress distribution, compromising their integrity. Detecting and repairing these cracks is, therefore, a priority in concrete engineering.

Traditionally, crack detection underwater relies on divers performing visual inspections and physical probing. These manual methods, however, face several limitations: they often produce fragmented data, miss certain areas, operate at low efficiency, and involve significant safety risks. Furthermore, inspection depth is constrained by the limits of diving operations, restricting access to deeper sections of structures.

With advancements in underwater robotics, cameras mounted on robotic systems have increasingly been used for inspection. Yet, underwater imaging remains challenging due to poor image quality, light scattering, and color distortion, which limit the effectiveness of these inspections.

To address these issues, this study proposes a non-contact method for detecting underwater concrete cracks. By combining monocular vision with fractal science and enhanced image-processing techniques, this approach aims to improve crack detection accuracy and provide a more reliable assessment of underwater structural health.

Methods

To start, the researchers created a baseline dataset of concrete cracks by conducting axial compression tests. Four primary concrete samples and a set of standard cubic specimens were prepared for these tests. The loading process was broken into two stages: an initial preloading phase at 50 KN to simulate lower stress and a final loading phase at 3000 KN to push the concrete to its limits. This setup allowed the team to capture realistic crack data for further analysis.

With this baseline in hand, they simulated an underwater environment to develop a dataset for underwater crack detection. To handle the challenges of underwater imaging, they built an image enhancement algorithm using an encoder structure. This enhancement was key to improving both the crack detection and segmentation needed in the underwater environment.

For accurate crack measurement, the team created mathematical models that mapped points between two-dimensional image coordinates and three-dimensional space. This step was essential for converting the camera images into precise spatial measurements, giving a more realistic view of the cracks.

Different segmentation techniques were then applied based on the crack type. For fine, narrow cracks, the team used the Prewitt operator to get clean segmentation. For wider cracks—often the ones posing more risk to structural integrity—they applied an optimized version of the Otsu method, which does a good job of highlighting significant crack areas.

Gamma correction was applied to reduce image brightness to improve detection even further, while a decoder-based restoration network helped correct color shifts caused by underwater light distortion. This process improved image contrast and clarity, making the cracks easier to detect and measure accurately. They also used convolution techniques to extract high-frequency details from the images, which were then represented in three modules: shrinkage, nonlinear mapping, and expansion.

Finally, to classify cracks based on their complexity, the team turned to fractal theory. Using MATLAB’s Fraclab toolbox, they calculated the fractal dimension of each crack, giving them a clear way to assess and categorize the severity of the damage.

Results and Discussion

The analysis showed that water turbidity had a significant impact on both crack measurement accuracy and the effective measurement range for underwater concrete structures. As turbidity increased, it became harder to accurately measure cracks from a distance. However, by applying underwater filtering, noise reduction, and image enhancement techniques, the team was able to improve crack recognition and restore visual details that helped identify and measure the cracks more effectively.

The accuracy of this method varied based on the level of water turbidity. In light turbidity conditions, the margin of error for crack width measurement was approximately 2 % at a distance of 0.5 m, 9 % at 0.8 m, and 16 % at 1.2 m.

In contrast, for heavy turbidity, the error increased to about 10 % at 0.5 m, 16 % at 0.8 m, and 28 % at 1.2 m. These results suggest that the system performs best at closer distances, particularly in clearer water conditions, making it well-suited for close-range inspections in turbid environments.

One interesting finding was the relationship between turbidity and the fractal dimension of the detected cracks. As turbidity levels rose, so did the fractal dimension, which negatively affected detection accuracy. Despite this, the fractal dimension still served as a helpful safety indicator, providing insight into the structure’s condition when visibility was limited.

While the system was unable to offer a complete safety assessment under highly turbid conditions, it could still act as a useful tool for structural monitoring and preliminary analysis.

Conclusion

Overall, the researchers proposed a fractal theory-based image-processing approach for detecting and measuring cracks in underwater concrete structures. By simulating an underwater environment, the researchers demonstrated the effectiveness of this computer vision-based method, particularly when combined with image enhancement techniques that improve crack visibility at short distances.

The algorithm showed promising results for detecting cracks in turbid water, proving practical and effective in specific conditions. This system offers an efficient solution for visualizing and monitoring underwater structures. However, further advancements are needed to extend its accuracy and effectiveness for longer-range imaging.

Journal Reference

Cui, B., Wang, C., Li, Y., Li, H., & Li, C. (2024). Application of computer vision techniques to damage detection in underwater concrete structures. Alexandria Engineering Journal, 104, 745–752. DOI: 10.1016/j.aej.2024.08.020, https://www.sciencedirect.com/science/article/pii/S1110016824009050

Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.