Deep learning-based segmentation and detection of tunnel lining defects and components from GPR images using T-GPRMask

Abstract

Ground penetrating radar (GPR) has been extensively applied in tunnel engineering for the non-destructive assessment of lining structures. However, the interpretation of GPR images remains a time-consuming and expertise-dependent task. To address this challenge, this study proposes tunnel ground-penetrating radar mask region-based convolutional neural network (T-GPRMask), a deep learning-based instance segmentation model designed for the automated detection of tunnel lining defects and components. By integrating a convolutional block attention module (CBAM) and feature pyramid network (FPN), T-GPRMask enhances multi-scale feature extraction, enabling the detection of small, low-contrast defects that are commonly encountered in GPR images. The model was pretrained on a domain-specific dataset containing a diverse set of GPR images related to underground structures and then fine-tuned on a dataset specifically designed for tunnel inspections. The model achieved recognition accuracies of 83.18%, 88.24%, 92.84%, and 91.56% for detecting poor compactness, voids, steel arch supports, and initial lining thickness, respectively. A comparative study further demonstrated T-GPRMask’s superior performance over traditional models, such as YOLOv7 and RetinaNet. Field experiments on real-world tunnel inspection data validated the model’s high spatial accuracy and highlighted its practical applicability in tunnel maintenance.

Keywords

Ground penetrating radar / Tunnel lining inspection / Instance segmentation / Deep learning / Transfer learning

Data availability

The data that used in this study are available on https://github.com/ljhtj97/T-GPRMask.

CRediT authorship contribution statement

Jiahao Li: Writing - original draft, Visualization, Validation, Methodology, Data curation. Hehua Zhu: Writing - review & editing, Validation, Supervision, Project administration, Funding acquisition. Mei Yin: Writing - review & editing, Investigation, Data curation.

Declaration of competing interest

Professor Hehua Zhu is an editor-in-chief for Underground Space and was not involved in the editorial review or the decision to publish this article. All authors declare that there are no competing interests.

Acknowledgement

This research is supported by the National Key Research and Development Program of China (Grant No. 2023YFC3009400).

References

[1] Afshani, A., Kawakami, K., Konishi, S., & Akagi, H. (2019). Study of infrared thermal application for detecting defects within tunnel lining. Tunnelling and Underground Space Technology, 86, 186-197.

[2] Borgioli, G., Capineri, L., Falorni, P., Matucci, S., & Windsor, C. G. (2008). The detection of buried pipes from time-of-flight radar data. IEEE Transactions on Geoscience and Remote Sensing, 46(8), 2254-2266.

[3] Capineri, L., Grande, P., & Temple, J. A. G. (1998). Advanced image-processing technique for real-time interpretation of ground-penetrating radar images. International Journal of Imaging Systems and Technology, 9(1), 51-59.

[4] De Coster, A., Medina, J. P., Nottebaere, M., Alkhalifeh, K., Neyt, X., Vanderdonckt, J., & Lambot, S. (2019). Towards an improvement of GPR-based detection of pipes and leaks in water distribution networks. Journal of Applied Geophysics, 162, 138-151.

[5] Dou, Q. X., Wei, L. J., Magee, D. R., & Cohn, A. G. (2016). Real-time hyperbola recognition and fitting in GPR data. IEEE Transactions on Geoscience and Remote Sensing, 55(1), 51-62.

[6] Girshick, R. (2015). Fast R-CNN. In Proceedings of 2015 IEEE International Conference on Computer Vision (ICCV) (pp. 1440-1448).

[7] Gong, C. J., Xie, C. R., Zhu, H. H., Ding, W. Q., Song, J. R., & Ge, Y. Y. (2024). Time-varying compressive properties and constitutive model of EPDM rubber materials for tunnel gasketed joint. Construction and Building Materials, 433, 136734.

[8] Gong, Z. M., & Zhang, H. Q. (2020). Research on GPR image recognition based on deep learning. MATEC Web of Conferences, 309, 03027.

[9] Guo, L., Cai, L. H., & Chen, D. J. (2022). Research on tunnel lining image target recognition method based on YOLOv3. In 2022 IEEE 10th Joint International Information Technology and Artificial Intelligence Conference (ITAIC) (pp. 1800-1805).

[10] Guo, S. L., Deng, J., Li, W. W., & Zhang, X. Q. (2019). Ground penetrating radar multi-profile integrated interpretation method and application. Progress in Geophysics, 34(5), 2022-2029.

[11] He, K. M., Gkioxari, G., Dollár, P., & Girshick, R. (2017). Mask R-CNN. In Proceedings of the IEEE International Conference on Computer Vision (ICCV) (pp. 2961-2969).

[12] Hoarau, Q., Ginolhac, G., Atto, A. M., & Nicolas, J. M. (2017). Robust adaptive detection of buried pipes using GPR. Signal Processing, 132, 293-305.

[13] Jin, X., Qiao, K. K., Bu, M. H., Wang, J. M., Wang, M., & Fang, C. (2025). Intelligent safety evaluation of tunnel lining cracks based on machine learning. Engineering Failure Analysis, 167, 109082.

[14] Lai, W. W. L., Dérobert, X., & Annan, P. (2018). A review of Ground Penetrating Radar application in civil engineering: A 30-year journey from locating and testing to imaging and diagnosis. NDT & E International, 96, 58-78.

[15] Lee, K. L., & Mokji, M. M. (2014). Automatic target detection in GPR images using Histogram of Oriented Gradients (HOG). In 2014 2nd International Conference on Electronic Design (ICED) (pp. 181-186).

[16] Lei, W. T., Hou, F. F., Xi, J. C., Tan, Q. Y., Xu, M. D., Jiang, X. Y., Liu, G. Y., & Gu, Q. Y. (2019). Automatic hyperbola detection and fitting in GPR B-scan image. Automation in Construction, 106, 102839.

[17] Li, S. W., Gu, X. Y., Xu, X. R., Xu, D. W., Zhang, T. J., Liu, Z., & Dong, Q. (2021). Detection of concealed cracks from ground penetrating radar images based on deep learning algorithm. Construction and Building Materials, 273, 121949.

[18] Li, W. T., Cui, X. H., Guo, L., Chen, J., Chen, X. H., & Cao, X. (2016). Tree root automatic recognition in ground penetrating radar profiles based on randomized Hough transform. Remote Sensing, 8(5), 430.

[19] Lin, T. Y., Dollár, P., Girshick, R., He, K. M., Hariharan, B., & Belongie, S. (2017a). Feature pyramid networks for object detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 936-944).

[20] Lin, T. Y., Goyal, P., Girshick, R., He, K. M., & Dollár, P. (2017b). Focal loss for dense object detection. In Proceedings of the 2017 IEEE International Conference on Computer Vision (ICCV) (pp. 2999-3007).

[21] Liu, H., Shi, Z. S., Li, J. H., Liu, C., Meng, X., Du, Y. L., & Chen, J. (2021). Detection of road cavities in urban cities by 3D ground-penetrating radar. Geophysics, 86(3), WA25-WA33.

[22] Liu, Z., Gu, X. Y., Chen, J. Q., Wang, D. Y., Chen, Y. H., & Wang, L. T. (2023). Automatic recognition of pavement cracks from combined GPR B-scan and C-scan images using multiscale feature fusion deep neural networks. Automation in Construction, 146, 104698.

[23] Loupos, K., Doulamis, A. D., Stentoumis, C., Protopapadakis, E., Makantasis, K., Doulamis, N. D., Amditis, A., Chrobocinski, P., Victores, J., Montero, R., Menendez, E., Balaguer, C., Lopez, R., Cantero, S., Navarro, R., Roncaglia, A., Belsito, L., Camarinopoulos, S., Komodakis, N., & Singh, P. (2018). Autonomous robotic system for tunnel structural inspection and assessment. International Journal of Intelligent Robotics and Applications, 2, 43-66.

[24] Maas, C., & Schmalzl, J. (2013). Using pattern recognition to automatically localize reflection hyperbolas in data from ground penetrating radar. Computers & Geosciences, 58, 116-125.

[25] Mertens, L., Persico, R., Matera, L., & Lambot, S. (2016). Automated detection of reflection hyperbolas in complex GPR images with no a priori knowledge on the medium. IEEE Transactions on Geoscience and Remote Sensing, 54(1), 580-596.

[26] Mojahid, A., El Ouai, D., El Amraoui, K., El-Hami, K., & Aitbenamer, H. (2025). Intelligent recognition of subsurface utilities and voids: A ground penetrating radar dataset for deep learning applications. Data in Brief, 59, 111338.

[27] Pan, Z. T., Zhang, X. P., Jiang, Y. J., Li, B., Golsanami, N., Su, H., & Cai, Y. (2025). High-precision segmentation and quantification of tunnel lining crack using an improved DeepLabV3+. Underground Space, 22, 96-109.

[28] Park, S., Kim, J., Jeon, K., Kim, J., & Park, S. (2021). Improvement of GPR-based rebar diameter estimation using YOLO-v3. Remote Sensing, 13(10), 2011.

[29] Pham, M. T., & Lefèvre, S. (2018). Buried object detection from B-scan ground penetrating radar data using Faster-RCNN. In IGARSS 2018-2018 IEEE International Geoscience and Remote Sensing Symposium (pp. 6804-6807). IEEE.

[30] Qin, H., Xie, X. Y., Tang, Y., & Wang, Z. Z. (2020). Experimental study on GPR detection of voids inside and behind tunnel linings. Journal of Environmental and Engineering Geophysics, 25(1), 65-74.

[31] Qin, H., Zhang, D. H., Tang, Y., & Wang, Y. Z. (2021). Automatic recognition of tunnel lining elements from GPR images using deep convolutional networks with data augmentation. Automation in Construction, 130, 103830.

[32] Qiu, Z., Zhao, Z. X., Chen, S. J., Zeng, J. Y., Huang, Y., & Xiang, B. R. (2022). Application of an improved YOLOv5 algorithm in real-time detection of foreign objects by ground penetrating radar. Remote Sensing, 14(8), 1895.

[33] Ren, S. Q., He, K. M., Girshick, R., & Sun, J. (2017). Faster R-CNN: Towards real-time object detection with region proposal networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 39(6), 1137-1149.

[34] Wang, C. Y., Bochkovskiy, A., & Liao, H. Y. M. (2023). YOLOv7:Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 7464-7475).

[35] Wang, J., Zhang, J. Q., Cohn, A. G., Wang, Z. F., Liu, H. C., Kang, W. Q., Jiang, P., Zhang, F. K., Chen, K. F., Guo, W., & Yu, Y. F. (2022). Arbitrarily-oriented tunnel lining defects detection from ground penetrating radar images using deep convolutional neural networks. Automation in Construction, 133, 104044.

[36] Woo, S., Park, J., Lee, J. Y., & Kweon, I. S. (2018). CBAM: Convolutional block attention module. In Proceedings of the 15th European Conference on Computer Vision (pp. 3-19).

[37] Youn, H. S., & Chen, C. C. (2002). Automatic GPR target detection and clutter reduction using neural network. In Ninth International Conference on Ground Penetrating Radar (GPR 2002) (pp. 579-582). SPIE.

[38] Yue, Y. P., Liu, H., Lin, C. D., Meng, X., Liu, C., Zhang, X. Y., Cui, J., & Du, Y. L. (2024). Automatic recognition of defects behind railway tunnel linings in GPR images using transfer learning. Measurement, 224, 113903.

[39] Zhou, Z., Zhang, J. J., Gong, C. J., & Wu, W. (2023). Automatic tunnel lining crack detection via deep learning with generative adversarial network-based data augmentation. Underground Space, 9, 140-154.