daohang fenxiangbox searchbox qikanlogonew daohangnew searchboxnew navrightzone footerzone paper
中
Home > Journals > Chain
‹ Previous Article Next Article ›

MARD-Net: fusing multi-path attention, re-parameterization, and asymmetric detection head for subsurface road defect detection in GPR imagery AITranslate

Wenbo Zhang1; Yi Liang1; Jueqiang Tao1; Qing Yang1; Zican Liu2; Chenyang Li3;
1.Road and Traffic Engineering Research Center, Zhejiang Normal University, Jinhua 321004, China
2.College of Engineering, Zhejiang Normal University, Jinhua 321004, China
3.College of Environment, Zhejiang University of Technology, Hangzhou 310023, China
AITranslate
Publisher: Youke Publish Co., Ltd.
Share Citation Information Add to Favorites Download PDF

    Scan to share on WeChat or Moments

Use WeChat scan.
Share with WeChat friends or Moments

Abstract AITranslate

Timely detection of subsurface road defects is critical for structural safety and pavement longevity. Here, we propose MARD-Net, an enhanced deep learning framework designed for the accurate identification of urban subsurface road defects. First, addressing the scarcity of defect samples, a hybrid dataset was constructed by integrating empirical data acquired via the GS8000 ground-penetrating radar (GPR) system and synthetic data generated by gprMax. Second, to address complex geological backgrounds and variations in defect waveform scales, the RepNCSPELAN4_CAA module was integrated into the architecture. By combining structural re-parameterization with context anchor attention (CAA), this module enhances feature reuse and inter-channel interaction, enabling the fine-grained discrimination of subtle defects. Third, a lightweight asymmetric detection head (LADH), incorporating depthwise separable convolution (DSConv) within its regression branch, was developed to significantly reduce computational costs while maintaining robust detection performance. Finally, to overcome weak and uneven defect signals across imaging depths, a multi-path coordinate attention (MPCA) mechanism adaptively fuses global and local contextual information for precise defect recognition. Empirical experiments show that MARD-Net achieves a 2.07% increase in mean average precision at an intersection over union threshold of 0.50 (mAP@50) over the baseline YOLOv11n while reducing floating point operations (FLOPs) by 2.0 giga floating point operations (GFLOPs).

KeyWords AITranslate

LADH MARD-Net MPCA RepNCSPELAN4_CAA subsurface road defects

1.X. C. Sui, Z. Leng, and S. Wang, "Machine Learning-Based Detection of Transportation Infrastructure Internal Defects Using Ground-Penetrating Radar: A State-of-the-Art Review," Intelligent Transportation Infrastructure 2, no. 1 (2023): 1–18, https://doi.org/10.1093/iti/liad004.

2.Z. Zou, K. Chen, Z. Shi, Y. Guo, and J. Ye, "Object Detection in 20 Years: A Survey," Proceedings of the IEEE 111, no. 3 (2023): 257–276, https://doi.org/10.1109/JPROC.2023.3238524.

3.R. Varghese and M. Sambath, "A Comprehensive Review on Two-stage Object Detection Algorithms," paper presented at the 2023 International Conference on Quantum Technologies, Communications, Computing, Hardware and Embedded Systems Security (iQ-CCHESS), Kottayam, India, September 15–16, 2023, https://doi.org/10.1109/iQ-CCHESS56596.2023.10391506.

4.M. T. Pham and S. Lefèvre, "Buried Object Detection from B-Scan Ground Penetrating Radar Data Using Faster-RCNN," paper presented at the IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Valencia, Spain, July 22–27, 2018, https://doi.org/10.1109/IGARSS.2018.8517683.

5.X. Xu, Y. Lei, and F. Yang, "Railway Subgrade Defect Automatic Recognition Method Based on Improved Faster R-CNN," Scientific Programming 2018, no. 2 (2018): 1–12, https://doi.org/10.1155/2018/4832972.

6.F. Niu, Y. Huang, P. He, Q. Wu, and Y. Zhang, "Intelligent Recognition of Ground Penetrating Radar Images in Urban Road Detection: A Deep Learning Approach," Journal of Civil Structural Health Monitoring 14, no. 5 (2024): 1917–1933, https://doi.org/10.1007/s13349-024-00818-5.

7.J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, "You Only Look Once: Unified, Real-Time Object Detection," paper presented at the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, June 27–30, 2016, https://doi.org/10.1109/CVPR.2016.91.

8.Z. K. Ni, D. Zhao, S. B. Ye, and G. Y. Fang, "City Road Cavity Detection Using YOLOv3 for Ground-Penetrating Radar," paper presented at the 18th International Conference on Ground Penetrating Radar, Golden, Colorado, June 14–19, 2020, https://doi.org/10.1190/gpr2020-100.1.

9.R. Mehta, P. K. R. Rao, M. Raj, and A. Khosla, "CNN-Based Sub-Surface Object Detection Using Ground Penetrating Radar," paper presented at the 11th International Workshop on Advanced Ground Penetrating Radar (IWAGPR), Valletta, Malta, December 1–4, 2021, https://doi.org/10.1109/IWAGPR50767.2021.9843163.

10.R. M. Hu, X. Li, X. Jing, J. Q. Wu, and Q. B. Wei, "Application of YOLOv7 in GPR B-Scan Image Interpretation," Bulletin of Surveying and Mapping 2023, no. 8 (2023): 29–33, http://tb.chinasmp.com/CN/Y2023/V0/I8/29.

11.D. S. Feng and Z. L. Yang, "Automatic Recognition of Ground Penetrating Radar Image of Tunnel Lining Structure Based on Deep Learning," Progress in Geophysics 35, no. 4 (2020): 1552–1556, https://doi.org/10.6038/pg2020DD0325.

12.C. Yi, J. Liu, T. Huang, H. Xiao, and H. Guan, "An Efficient Method of Pavement Distress Detection Based on Improved YOLOv7," Measurement Science and Technology 34, no. 11 (2023): 115402, https://doi.org/10.1088/1361-6501/ace929.

13.T. T. Fang, C. S. Wang, J. Wang, and Y. B. Du, "Pipeline Location Method of Ground-Penetrating Radar Images Based on YOLOv8n," Foreign Electronic Measurement Technology 42, no. 11 (2023): 170–177, https://doi.org/10.19652/j.cnki.femt.2305179.

14.W. Wang, "Advanced Auto Labeling Solution with Added Features," GitHub repository, 2023, https://github.com/CVHub520/X-AnyLabeling.

15.Ministry of Housing and Urban-Rural Development of the P.R.C. Technical Standard for Comprehensive Detection and Risk Assessment of Urban Underground Defects. China Architecture & Building Press JGJ/T 437-2018. 2018.

16.Ultralytics, "Ultralytics YOLO11," GitHub repository, 2024, https://github.com/ultralytics/ultralytics.

17.M. Yaseen, "What Is YOLOv8: An In-Depth Exploration of the Internal Features of the Next-Generation Object Detector," arXiv preprint, arXiv: 2408.15857, August 28, 2024, https://doi.org/10.48550/arXiv.2408.15857.

18.Z. Zhou, A. He, Y. Wu, R. Yao, X. Xie, and T. Li, "Spatial-Frequency Dual Domain Attention Network for Medical Image Segmentation," paper presented at the 2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), Lisbon, Portugal, December 3–6, 2024, https://doi.org/10.1109/BIBM62325.2024.10822613.

19.J. Zhang, Z. Chen, G. Yan, Y. Wang, and B. Hu, "Faster and Lightweight: An Improved YOLOv5 Object Detector for Remote Sensing Images," Remote Sensing 15, no. 20 (2023): 4974, https://doi.org/10.3390/rs15204974.

20.X. Cai, Q. Lai, Y. Wang, W. Wang, Z. Sun, and Y. Yao, "Poly Kernel Inception Network for Remote Sensing Detection," paper presented at the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, June 17–21, 2024, https://doi.org/10.48550/arXiv.2403.06258.

21.S. Ren, K. He, R. Girshick, and J. Sun, "Faster R-CNN: Towards Real-time Object Detection with Region Proposal Networks," IEEE Transactions on Pattern Analysis and Machine Intelligence 39, no. 6 (2017): 1137–1149, https://doi.org/10.1109/TPAMI.2016.2577031.

22.Z. Cai and N. Vasconcelos, "Cascade R-CNN: Delving into High Quality Object Detection," paper presented at the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Salt Lake City, UT, USA, June 18–22, 2018, https://doi.org/10.1109/CVPR.2018.00644.

23.H. Zhang, F. Li, S. Liu, et al., "DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection," paper presented at the International Conference on Learning Representations (ICLR), Kigali, Rwanda, May 1–5, 2023, https://doi.org/10.48550/arXiv.2203.03605.

24.C. Feng, Y. Zhong, Y. Gao, M. R. Scott, and W. Huang, "TOOD: Task-Aligned One-Stage Object Detection," paper presented at the IEEE/CVF International Conference on Computer Vision (ICCV), Montreal, QC, Canada, October 11–17, 2021, https://doi.org/10.1109/ICCV48922.2021.00349.

25.T. Y. Lin, P. Goyal, R. Girshick, K. He, and P. Dollár, "Focal Loss for Dense Object Detection," paper presented at the IEEE International Conference on Computer Vision (ICCV), Venice, Italy, October 22–29, 2017, https://doi.org/10.1109/ICCV.2017.324.

26.M. Sohan, T. Sai Ram, and C. V. Rami Reddy, "A Review on YOLOv8 and Its Advancements," in proceedings of the Data Intelligence and Cognitive Informatics, edited by I. J. Jacob, S. Piramuthu, and P. Falkowski-Gilski, 491–502, Springer, 2024, https://doi.org/10.1007/978-981-99-7962-2_39.

27.A. Wang, H. Chen, L. Liu, et al., "YOLOv10: Real-Time End-to-End Object Detection," paper presented at the Conference on Neural Information Processing Systems (NeurIPS), Vancouver, BC, Canada, December 10–15, 2024, https://doi.org/10.48550/arXiv.2405.14458.

28.R. Khanam and M. Hussain, "YOLOv11: An Overview of the Key Architectural Enhancements," arXiv preprint, arXiv: 2410.17725, October 23, 2024, https://doi.org/10.48550/arXiv.2410.17725.

29.Y. Tian, Q. Ye, and D. Doermann, "YOLOv12: Attention-Centric Real-Time Object Detectors," arXiv preprint, arXiv: 2502.12524, February 18, 2025, https://doi.org/10.48550/arXiv.2502.12524.

30.M. Tan, R. Pang, and Q. V. Le, "EfficientDet: Scalable and Efficient Object Detection," paper presented at the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, June 14–19, 2020, https://doi.org/10.1109/CVPR42600.2020.01079.

31.C. Y. Wang, I. H. Yeh, and H. Y. Mark Liao, "YOLOv9: Learning What You Want to Learn Using Programmable Gradient Information," paper presented at the European Conference on Computer Vision (ECCV), Milan, Italy, September 29 to October 4, 2024, https://doi.org/10.1007/978-3-031-72751-1_1.

32.J. B. Cordonnier, A. Loukas, and M. Jaggi, "Multi-Head Attention: Collaborate Instead of Concatenate," arXiv preprint, arXiv: 2006.16362, June 29, 2020, https://doi.org/10.48550/arXiv.2006.16362.

33.X. Li, W. Wang, X. Hu, J. Li, J. Tang, and J. Yang, "Generalized Focal Loss V2: Learning Reliable Localization Quality Estimation for Dense Object Detection," paper presented at the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Nashville, TN, USA, June 19–25, 2021, https://doi.org/10.1109/CVPR46437.2021.01146.

34.M. Hu, J. Feng, J. Hua, et al., "Online Convolutional Reparameterization," paper presented at the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), New Orleans, LA, USA, June 19–24, 2022, https://doi.org/10.1109/CVPR52688.2022.00065.

35.C. Li, A. Zhou, and A. Yao, "Omni-Dimensional Dynamic Convolution," paper presented at the International Conference on Learning Representations (ICLR), Virtual, April 25–29, 2022, https://doi.org/10.48550/arXiv.2209.07947.

36.X. Zhang, Y. Song, T. Song, et al., "AKConv: Convolutional Kernel with Arbitrary Sampled Shapes and Arbitrary Number of Parameters," arXiv preprint, arXiv: 2311.11587, November 20, 2023, https://doi.org/10.48550/arXiv.2311.11587.

37.Y. Chen, X. Dai, M. Liu, D. Chen, L. Yuan, and Z. Liu, "Dynamic Convolution: Attention Over Convolution Kernels," paper presented at the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, June 14–19, 2020, https://doi.org/10.1109/CVPR42600.2020.01104.

38.D. Shi, "TransNeXt: Robust Foveal Visual Perception for Vision Transformers," paper presented at the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, June 17–21, 2024, https://doi.org/10.48550/arXiv.2311.17132.

39.L. Yang, R. Y. Zhang, L. Li, and X. Xie, "SimAM: A Simple, Parameter-Free Attention Module for Convolutional Neural Networks," paper presented at the International Conference on Machine Learning (ICML), Virtual, July 18–24, 2021, http://proceedings.mlr.press/v139/yang21o.html.

40.Q. Hou, D. Zhou, and J. Feng, "Coordinate Attention for Efficient Mobile Network Design," paper presented at the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Nashville, TN, USA, June 19–25, 2021, https://doi.org/10.1109/CVPR46437.2021.01350.

41.D. Wan, R. Lu, S. Shen, T. Xu, X. Lang, and Z. Ren, "Mixed Local Channel Attention for Object Detection," Engineering Applications of Artificial Intelligence 123, Part C (2023): 106442, https://doi.org/10.1016/j.engappai.2023.106442.

42.L. Zhu, X. Wang, Z. Ke, W. Zhang, and R. Lau, "BiFormer: Vision Transformer with Bi-Level Routing Attention," paper presented at the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Vancouver, BC, Canada, June 18–22, 2023, https://doi.org/10.48550/arXiv.2303.08810.

Basic Information:

DOI：10.23919/CHAIN.2026.000003

Chinese Library Classification Number:

Citation Information:

Timely detection of subsurface road defects is critical for structural safety and pavement longevity. Here, we propose MARD-Net, an enhanced deep learning framework designed for the accurate identification of urban subsurface road defects. First, addressing the scarcity of defect samples, a hybrid dataset was constructed by integrating empirical data acquired via the GS8000 ground-penetrating radar (GPR) system and synthetic data generated by gprMax. Second, to address complex geological backgrounds and variations in defect waveform scales, the RepNCSPELAN4_CAA module was integrated into the architecture. By combining structural re-parameterization with context anchor attention (CAA), this module enhances feature reuse and inter-channel interaction, enabling the fine-grained discrimination of subtle defects. Third, a lightweight asymmetric detection head (LADH), incorporating depthwise separable convolution (DSConv) within its regression branch, was developed to significantly reduce computational costs while maintaining robust detection performance. Finally, to overcome weak and uneven defect signals across imaging depths, a multi-path coordinate attention (MPCA) mechanism adaptively fuses global and local contextual information for precise defect recognition. Empirical experiments show that MARD-Net achieves a 2.07% increase in mean average precision at an intersection over union threshold of 0.50 (mAP@50) over the baseline YOLOv11n while reducing floating point operations (FLOPs) by 2.0 giga floating point operations (GFLOPs).

quote

GB/T 7714-2015 [1] Wenbo Zhang, Yi Liang, Jueqiang Tao, et al. MARD-Net: fusing multi-path attention, re-parameterization, and asymmetric detection head for subsurface road defect detection in GPR imagery[J]. Chain, 2026, 3(1): 94-116. DOI:10.23919/CHAIN.2026.000003.
MLA [1] Wenbo Zhang, et al., "MARD-Net: fusing multi-path attention, re-parameterization, and asymmetric detection head for subsurface road defect detection in GPR imagery." Chain, vol. 3, no. 1, 2026, pp. 94-116, https://doi.org/10.23919/CHAIN.2026.000003.
APA [1] Wenbo Zhang, Yi Liang, Jueqiang Tao, Qing Yang, Zican Liu, & Chenyang Li. (2026). MARD-Net: fusing multi-path attention, re-parameterization, and asymmetric detection head for subsurface road defect detection in GPR imagery. Chain, 3(1), 94-116. https://doi.org/10.23919/CHAIN.2026.000003
IEEE [1] Wenbo Zhang, Yi Liang, Jueqiang Tao, Qing Yang, Zican Liu, and Chenyang Li, "MARD-Net: fusing multi-path attention, re-parameterization, and asymmetric detection head for subsurface road defect detection in GPR imagery," Chain, vol. 3, no. 1, pp. 94-116, 2026, doi: 10.23919/CHAIN.2026.000003. keywords: {LADH;MARD-Net;MPCA;RepNCSPELAN4_CAA;subsurface road defects}