Automatic detection of large-scale pavement layer debonding via light-weight deep multi-view learning model and B-scan composite images
Source abstract
Abstract
Large-scale interlayer debonding on road surfaces typically extends across multiple channels in ground-penetrating radar (GPR) B-scan images. Existing single-view automatic detection models are time-consuming and ineffective in capturing correlated features between adjacent channels, which are essential for accurate identification of this defect. To address this challenge, this study proposes a large-scale debonding detection method based on a deep multi-view learning model and composite B-scan images. First, a multi-view B-scan composite image was generated using a spatial stitching method, allowing the spatial representation of large-scale debonding. Next, a lightweight multi-view-one classification network named DB-mv-cls was proposed to locate debonding area, achieving accuracy of 83% for debondings, 88% for normal areas. This model demonstrated the best detection performance among 20 models, with detection efficiency for 10-channel images approximately four times greater than that of the single-view detection mode. The method enhances automated detection for large-scale debonding.
Source: openalex_inverted_index · Quality: baseline
Research context
OpenAlex topics: Geophysical Methods and Applications · Infrastructure Maintenance and Monitoring · Asphalt Pavement Performance Evaluation
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
Affiliations
Chang’an University
School of Future Transportation, Chang’an University
Pingdu City Transportation Bureau
Zhongjiao Yuanzhou Transportation Technology Group Shandong Engineering Consulting Co., Ltd
Funding
National Natural Science Foundation of China · 111 project of Sustainable Development of Transportation in Western Urban Agglomeration · 111 project of Low carbon and intelligent road infrastructure and Innovative maintenance discipline · Fundamental Research Funds for the Central Universities, CHD · 2025 Chang’an University Graduate Research Practice Innovation Project
References (81)
Show source-provided references
1. Rahma K, Anne M, Fazia F et al. Experimental characterization of the interface damage between bituminous layers: from the interphase to the interface properties. In: 8th RILEM International Conference on Mechanisms of Cracking and Debonding in Pavements (MCD); Nantes, FRANCE; 2016. p. 581–587. doi: 10.1007/978–94–024–0867–6_82
2. Ktari R, Fouchal F, Millien A et al. Damage modeling of asphaltic pavement rough interfaces under tensile and shear loading. In: 8th RILEM International Conference on Mechanisms of Cracking and Debonding in Pavements (MCD); Nantes, FRANCE; 2016. p. 475–481. doi: 10.1007/978-94-024-0867-6_66
3. Hernando D, Magruder JA, Zou J et al. Localized debonding as a potential mechanism for near-surface cracking. In: 8th RILEM International Conference on Mechanisms of Cracking and Debonding in Pavements (MCD); Nantes, FRANCE; 2016. p. 569–574. doi: 10.1007/978-94-024-0867-6_80
4. Grellet D, Dore G, Chupin O et al. Experimental evidence of the viscoelastic behavior of interfaces in bituminous pavements-an explanation to top-down cracking? In: 8th RILEM International Conference on Mechanisms of Cracking and Debonding in Pavements (MCD); Nantes, FRANCE; 2016. p. 575–580. doi: 10.1007/978-94-024-0867-6_81
5. 10.1080/19648189.2017.132023710.1080/19648189.2017.1320237 ↗
6. 10.1080/19648189.2017.131180810.1080/19648189.2017.1311808 ↗
7. Apeagyei AK, Grenfell JRA, Airey GD. Moisture-induced debonding mechanisms in asphalt mixtures. In: 8th RILEM International Conference on Mechanisms of Cracking and Debonding in Pavements (MCD); Nantes, FRANCE; 2016. p. 589–595. doi: 10.1007/978-94-024-0867-6_83
8. 10.1617/s11527-018-1223-y10.1617/s11527-018-1223-y ↗
9. 10.1016/j.conbuildmat.2022.12943410.1016/j.conbuildmat.2022.129434 ↗
10. 10.1016/j.conbuildmat.2023.13283410.1016/j.conbuildmat.2023.132834 ↗
11. 10.1016/j.conbuildmat.2018.02.10010.1016/j.conbuildmat.2018.02.100 ↗
12. 10.1016/j.autcon.2024.10533510.1016/j.autcon.2024.105335 ↗
13. 10.1016/j.conbuildmat.2023.13156410.1016/j.conbuildmat.2023.131564 ↗
14. 10.1016/j.conbuildmat.2022.12958810.1016/j.conbuildmat.2022.129588 ↗
15. 10.1016/j.autcon.2023.10502110.1016/j.autcon.2023.105021 ↗
16. 10.1016/j.conbuildmat.2024.13720910.1016/j.conbuildmat.2024.137209 ↗
17. 10.1016/j.conbuildmat.2022.12904210.1016/j.conbuildmat.2022.129042 ↗
18. 10.1016/j.conbuildmat.2023.13367410.1016/j.conbuildmat.2023.133674 ↗
19. 10.1016/j.autcon.2020.10311910.1016/j.autcon.2020.103119 ↗
20. 10.1016/j.conbuildmat.2024.13667510.1016/j.conbuildmat.2024.136675 ↗
21. 10.1016/j.conbuildmat.2021.12351810.1016/j.conbuildmat.2021.123518 ↗
22. 10.1016/j.conbuildmat.2022.12883110.1016/j.conbuildmat.2022.128831 ↗
23. 10.1016/j.measurement.2023.11329610.1016/j.measurement.2023.113296 ↗
24. 10.1016/j.autcon.2022.10469810.1016/j.autcon.2022.104698 ↗
25. 10.1016/j.jappgeo.2021.10439210.1016/j.jappgeo.2021.104392 ↗
26. Simonin JM, Baltazart V, Hornych P et al. Case study of detection of artificial defects in an experimental pavement structure using 3D OPR systems. In: 15th International Conference on Ground Penetrating Radar (GPR); Brussels, BELGIUM; 2014. p. 847–851.10.1109/ICGPR.2014.6970547 ↗
27. 10.1016/j.jappgeo.2022.10472210.1016/j.jappgeo.2022.104722 ↗
28. 10.1016/j.ndteint.2020.10229310.1016/j.ndteint.2020.102293 ↗
29. 10.1016/j.measurement.2020.10766210.1016/j.measurement.2020.107662 ↗
30. 10.1080/10298436.2020.177869210.1080/10298436.2020.1778692 ↗
31. 10.1016/j.autcon.2022.10468910.1016/j.autcon.2022.104689 ↗
32. 10.1016/j.autcon.2022.10451610.1016/j.autcon.2022.104516 ↗
33. 10.1016/j.conbuildmat.2020.12194910.1016/j.conbuildmat.2020.121949 ↗
34. 10.1016/j.conbuildmat.2019.11735210.1016/j.conbuildmat.2019.117352 ↗
35. 10.1016/j.conbuildmat.2017.04.09710.1016/j.conbuildmat.2017.04.097 ↗
36. 10.1016/j.autcon.2022.10426010.1016/j.autcon.2022.104260 ↗
37. 10.1109/tits.2023.331900310.1109/tits.2023.3319003 ↗
38. 10.1088/1361-6501/acb6e310.1088/1361-6501/acb6e3 ↗
39. 10.1109/tits.2022.322476910.1109/tits.2022.3224769 ↗
40. 10.1109/jstars.2022.316566010.1109/jstars.2022.3165660 ↗
41. 10.1016/j.conbuildmat.2023.13448310.1016/j.conbuildmat.2023.134483 ↗
42. 10.1007/s00521-013-1362-610.1007/s00521-013-1362-6 ↗
43. 10.1088/1361-6560/ad7f1b10.1088/1361-6560/ad7f1b ↗
44. 10.1016/j.patcog.2026.11369210.1016/j.patcog.2026.113692 ↗
45. 10.1016/j.patcog.2026.11343210.1016/j.patcog.2026.113432 ↗
46. 10.1016/j.neucom.2021.03.09010.1016/j.neucom.2021.03.090 ↗
47. 10.1016/j.patcog.2019.03.00410.1016/j.patcog.2019.03.004 ↗
48. 10.1109/tip.2020.299441210.1109/tip.2020.2994412 ↗
49. 10.1016/j.patcog.2019.05.03210.1016/j.patcog.2019.05.032 ↗
50. Ruijie H, Yanran L, Ningyu Z et al. Shifting perspective to see difference: a novel multi-view method for skeleton based action recognition. In: 30th ACM International Conference on Multimedia (MM); 2022.
51. Xu B, Li Y, Cao Q et al. Acm. In: Simple Multi-view Can Bring Powerful Graph Neural Network, 32nd World Wide Web Conference (WWW); Austin, TX; 2023. p. 322–325. doi: 10.1145/3543873.3587375
52. 10.1080/13658816.2023.220321810.1080/13658816.2023.2203218 ↗
53. Zhao L, Wang Z, Yuan Y et al. Unrestricted anchor graph based GCN for incomplete multi-view clustering. In: IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP); 2023.
54. 10.1371/journal.pone.018829010.1371/journal.pone.0188290 ↗
55. 10.1007/s12559-018-9598-110.1007/s12559-018-9598-1 ↗
56. Mane R, Robinson N, Vinod AP et al. Ieee, a multi-view CNN with novel variance layer for motor imagery brain computer interface. In: 42nd Annual International Conference of the IEEE-Engineering-in-Medicine-and-Biology-Society (EMBC); Montreal, CANADA; 2020. p. 2950–2953.
57. 10.1016/j.knosys.2026.11566010.1016/j.knosys.2026.115660 ↗
58. 10.3390/rs1807106310.3390/rs18071063 ↗
59. 10.1177/147592171983808110.1177/1475921719838081 ↗
60. 10.1177/147592172090270010.1177/1475921720902700 ↗
61. 10.3390/rs1418447210.3390/rs14184472 ↗
62. 10.1109/lra.2021.306259910.1109/lra.2021.3062599 ↗
63. 10.1109/TGRS.2023.330820510.1109/TGRS.2023.3308205 ↗
64. 10.1016/j.autcon.2023.10517910.1016/j.autcon.2023.105179 ↗
65. 10.1109/lgrs.2022.322878410.1109/lgrs.2022.3228784 ↗
66. 10.1080/10298436.2022.204627310.1080/10298436.2022.2046273 ↗
67. Yang F, Zhang QS, Wang PY. Research on detection technology of highway subgrade geological radar. Beijing: People’s Communications Press; 2009.
68. 10.1007/s40948-020-00147-z10.1007/s40948-020-00147-z ↗
69. 10.1016/j.conbuildmat.2021.12352110.1016/j.conbuildmat.2021.123521 ↗
70. Howard AG Zhu M Chen B et al. Mobilenets: efficient convolutional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861 (2017.
71. Ioannou Y, Robertson D, Cipolla R et al. Deep roots: improving CNN efficiency with hierarchical filter groups. In: Proceedings of the IEEE conference on computer vision and pattern recognition; 2017. p. 1231–1240.
72. Shen-Fu H, Bo-Ching T. Efficient computation of depthwise separable convolution in MobileNet deep neural network models. In: 2021 IEEE International Conference On Consumer Electronics-Taiwan, ICCE-TW; 2021.
73. Qararyah F, Azhar MW, Maleki MA et al. Assoc computing, fusing depthwise and pointwise convolutions for efficient inference on GPUs. In: 53rd International Conference on Parallel Processing (ICPP); Gotland, SWEDEN; 2024. p. 58–67. doi: 10.1145/3646548.3672597
74. Chun Q Jun-Gi J Hyun Dong L et al. Falcon: fast and lightweight convolution for compressing and accelerating CNN arXiv. arXiv. 2019;17.
75. 10.3390/agronomy1307182410.3390/agronomy13071824 ↗
76. 10.3390/rs1419480110.3390/rs14194801 ↗
77. 10.1111/mice.1283610.1111/mice.12836 ↗
78. 10.1109/lsp.2018.281012110.1109/lsp.2018.2810121 ↗
79. 10.1016/j.ecoinf.2020.10118210.1016/j.ecoinf.2020.101182 ↗
80. 10.3390/s2108285210.3390/s21082852 ↗
81. Liu Z, Lin Y, Cao Y et al. Ieee. In: Swin Transformer: Hierarchical Vision Transformer using Shifted Windows, 18th IEEE/CVF International Conference on Computer Vision (ICCV), Electr Network; 2021. p. 9992–10002. doi: 10.1109/iccv48922.2021.00986