Related Experiment Video
Updated: Dec 29, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
943
Automatic Tunnel Crack Detection Based on U-net and a Convolutional Neural Network with Alternately Updated Clique
Gang Li1,2, Biao Ma1, Shuanhai He3
1School of Electronic and Control Engineering, Chang'an University, Xi'an 710064, Shaanxi, China.
Sensors (Basel, Switzerland)
|February 5, 2020
Summary
This study introduces U-CliqueNet, a deep learning algorithm for accurate tunnel crack detection. The new method significantly improves crack segmentation compared to existing techniques, enhancing tunnel safety.
Area of Science:
- Civil Engineering
- Computer Science
- Artificial Intelligence
Background:
- Tunnel crack inspection is crucial for safety but current methods are inefficient or inaccurate.
- Detecting tiny cracks in noisy tunnel images presents a significant challenge.
Purpose of the Study:
- To develop a novel deep learning algorithm for precise tunnel crack semantic segmentation.
- To improve the accuracy and efficiency of automatic tunnel crack detection.
Main Methods:
- A hybrid deep learning model, U-CliqueNet, combining U-Net and CliqueNet architectures was proposed.
- A dataset of 60,000 tunnel crack sub-images was created for training and testing.
- Performance was evaluated using metrics like Mean Pixel Accuracy (MPA) and Mean Intersection over Union (MIoU) and compared against other algorithms.
Main Results:
- U-CliqueNet achieved high performance with MPA of 92.25% and MIoU of 86.96%.
- Hypothesis testing confirmed U-CliqueNet's significantly higher MIoU compared to FCN, U-net, SegNet, and MFCD.
- The framework accurately calculated crack dimensions with a mean crack width error ranging from -11.20% to 18.57%.
Conclusions:
- The U-CliqueNet framework offers a fast and accurate solution for tunnel crack semantic segmentation.
- This deep learning approach enhances the reliability of tunnel structural health monitoring.