Defect segmentation: Mapping tunnel lining internal defects with ground penetrating radar data using a convolutional neural network

Senlin Yang, Zhengfang Wang, Jing Wang, Anthony G. Cohn, Jiaqi Zhang, Peng Jiang, Lichao Nie, Qingmei Sui

TL;DR

Researchers have developed an automated "defect segmentation" method that uses a Convolutional Neural Network (CNN) to transform complex Ground Penetrating Radar (GPR) signals into clear, pixel-level maps of tunnel lining defects. By combining the SegNet architecture with a specialized Lovász softmax loss function, the system can accurately identify and shape tiny cracks and voids that traditional methods often miss.

What Problem Does This Paper Address?

Maintaining tunnel safety is a high-stakes engineering challenge. Tunnel linings often develop internal defects like cracks, voids, or "lining-rock separation" (gaps between the concrete and the surrounding earth) due to aging and geological stress.

Ground Penetrating Radar (GPR) is the gold standard for non-destructive testing because it’s fast and penetrates deep into structures. However, GPR doesn't produce a "photo" of the inside; it produces a "B-scan" filled with overlapping hyperbolic curves that represent reflected electromagnetic waves. Interpreting these scans currently requires highly trained experts, is prone to human error, and makes it difficult to determine the exact shape and location of a defect—especially when multiple defects or reinforcing rebars are present.

Methods

The authors propose a deep learning approach to automate this interpretation, treating it as a semantic segmentation task.

  1. Architecture: They chose SegNet, an encoder-decoder CNN. Unlike the popular U-Net, SegNet stores the indices of its max-pooling layers to perform non-linear upsampling. This helps the model maintain high-frequency structural integrity, which is vital for mapping sharp boundaries of internal defects.
  2. The "Secret Sauce" (Lovász Softmax Loss): Standard cross-entropy loss functions often fail when dealing with "imbalanced" data—for example, when a tiny crack (a few pixels) is surrounded by a massive block of concrete (thousands of pixels). The authors implemented the Lovász softmax loss, which optimizes the Mean Intersection over Union (MIoU). This significantly improved the model's ability to detect small, high-resolution features like rebars and thin cracks.
  3. Simulation to Reality: Because real-world labeled data of tunnel defects is scarce, the team used Finite Difference Time Domain (FDTD) simulations to generate a massive synthetic training set.

Datasets

  • Synthetic Dataset: The authors designed 124,800 sets of tunnel lining models. These models covered nine categories, including rebar, concrete, rock, and various types of defects (cracks, voids, separations) both with and without water.
  • Physical Test Model: To verify the results, they built a concrete testbed (4.4m x 2m x 0.7m) simulating a real tunnel lining. They embedded PVC pipes to create known voids and separations and collected real GPR data using a 600MHz antenna.

Key Insights and Ideas

  • Beyond Classification: Most previous GPR-AI research focused on identifying if a defect exists. This paper moves to segmentation, providing a pixel-by-pixel map of the defect's geometry.
  • The Rebar Shielding Effect: One of the most interesting "aha" moments is the quantification of how rebars interfere with signals. Rebars create a "shielding" effect that can hide or distort defects located beneath them. The paper shows that while the CNN can still classify these defects, the accuracy of their predicted shape drops significantly.
  • Loss Function Matters: The transition from a single loss (cross-entropy) to a composite loss (cross-entropy + Lovász) was the turning point for detecting thin cracks, which are often "washed out" by standard training methods.

Key Results

  • Superior Accuracy: The proposed SegNet with composite loss achieved a Mean Pixel Accuracy (MPA) of 93% and a Mean Intersection over Union (MIoU) of 90% on synthetic data.
  • Comparison: It significantly outperformed U-Net, which struggled with overfitting and failed to accurately classify complex overlapping signals (like separations and cracks occurring at different depths).
  • Real-World Validation: On the physical test model, the CNN correctly predicted the classification and location of water-bearing voids. While the predicted shape was less precise than in simulations due to real-world noise, the method proved its feasibility for field engineering.

Limitations and Future Work

  • Noise Sensitivity: Real-world GPR data is much noisier than FDTD simulations. The authors noted that the CNN needs to be "fine-tuned" with actual background noise from the specific environment to maintain high performance.
  • Complex Geometry: While the model handles flat and simple shapes well, very complex or irregular defect morphologies remain a challenge for precise shape reconstruction.
  • Future Direction: The authors suggest that future work should focus on collecting more diverse "background signals" from various real-world environments to create a more robust, "plug-and-play" model for different tunnel types.

Summary updated

Abstract

This research proposes a Ground Penetrating Radar (GPR) data processing method for non-destructive detection of tunnel lining internal defects, called defect segmentation. To perform this critical step of automatic tunnel lining detection, the method uses a CNN called Segnet combined with the Lov\'asz softmax loss function to map the internal defect structure with GPR synthetic data, which improves the accuracy, automation and efficiency of defects detection. The novel method we present overcomes several difficulties of traditional GPR data interpretation as demonstrated by an evaluation on both synthetic and real datas -- to verify the method on real data, a test model containing a known defect was designed and built and GPR data was obtained and analyzed.
BibTeX
@inproceedings{yang2021defect,
  author = {Senlin Yang and Zhengfang Wang and Jing Wang and Anthony G. Cohn and Jiaqi Zhang and Peng Jiang and Lichao Nie and Qingmei Sui},
  title = {{Defect segmentation: Mapping tunnel lining internal defects with ground penetrating radar data using a convolutional neural network}},
  booktitle = {Construction and Building Materials},
  year = {2021},
  url = {https://openreview.net/forum?id=MXurRCvQep},
}

Loading related content...

Figure