Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
139 tokens/sec
GPT-4o
47 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Deep Learning for Segmentation of Cracks in High-Resolution Images of Steel Bridges (2403.17725v1)

Published 26 Mar 2024 in cs.CV and eess.IV

Abstract: Automating the current bridge visual inspection practices using drones and image processing techniques is a prominent way to make these inspections more effective, robust, and less expensive. In this paper, we investigate the development of a novel deep-learning method for the detection of fatigue cracks in high-resolution images of steel bridges. First, we present a novel and challenging dataset comprising of images of cracks in steel bridges. Secondly, we integrate the ConvNext neural network with a previous state-of-the-art encoder-decoder network for crack segmentation. We study and report, the effects of the use of background patches on the network performance when applied to high-resolution images of cracks in steel bridges. Finally, we introduce a loss function that allows the use of more background patches for the training process, which yields a significant reduction in false positive rates.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (82)
  1. doi:https://doi.org/10.1016/j.jtte.2021.12.003.
  2. doi:https://doi.org/10.1080/15732479.2020.1832539.
  3. doi:https://doi.org/10.1061/(ASCE)BE.1943-5592.0001507.
  4. doi:https://doi.org/10.13023/KTC.RR.2016.26.
  5. doi:https://doi.org/10.3390/app11209757.
  6. doi:https://doi.org/10.3390/s20143954.
  7. doi:https://doi.org/10.1109/LRA.2019.2895880.
  8. doi:https://doi.org/10.3390/drones6110355.
  9. doi:https://doi.org/10.3390/app12031374.
  10. doi:https://doi.org/10.3390/buildings12040432.
  11. doi:10.1080/10298436.2018.1485917.
  12. doi:https://doi.org/10.1016/j.conbuildmat.2022.129659.
  13. doi:https://doi.org/10.3390/su14031825.
  14. doi:https://doi.org/10.1016/j.jtte.2022.11.003.
  15. doi:https://doi.org/10.3390/s19194251.
  16. doi:https://doi.org/10.1109/TIP.2018.2878966.
  17. doi:https://doi.org/10.1109/JSEN.2021.3089718.
  18. doi:https://doi.org/10.1109/JSEN.2019.2934897.
  19. doi:https://doi.org/10.1016/j.knosys.2022.108338.
  20. doi:https://doi.org/10.1109/TITS.2021.3106647.
  21. doi:https://doi.org/10.1109/TIM.2021.3075022.
  22. doi:https://doi.org/10.1109/TII.2020.3033170.
  23. doi:https://doi.org/10.1111/mice.12477.
  24. doi:https://doi.org/10.1111/mice.12844.
  25. doi:https://doi.org/10.1016/j.rineng.2023.101267.
  26. doi:https://doi.org/10.1061/(ASCE)CP.1943-5487.0000883.
  27. doi:https://doi.org/10.1109/ACCESS.2020.2980086.
  28. doi:https://doi.org/10.1109/TITS.2020.2990703.
  29. doi:https://doi.org/10.48550/arXiv.2010.11929.
  30. doi:https://doi.org/10.48550/arXiv.2304.02643.
  31. doi:https://doi.org/10.1016/j.autcon.2022.104646.
  32. doi:https://doi.org/10.1007/s00521-023-08277-7.
  33. doi:https://doi.org/10.1016/j.autcon.2022.104275.
  34. doi:https://doi.org/10.48550/arXiv.2105.15203.
  35. doi:https://doi.org/10.1007/s00138-020-01098-x.
  36. doi:https://doi.org/10.1007/s10851-023-01147-w.
  37. doi:https://doi.org/10.1016/j.dsp.2020.102907.
  38. doi:https://doi.org/10.1177/14759217211006485.
  39. doi:https://doi.org/10.1177/1475921718764873.
  40. doi:https://doi.org/10.1007/s13349-021-00537-1.
  41. doi:https://doi.org/10.21595/mrcm.2021.22032.
  42. doi:https://doi.org/10.3390/s21124135.
  43. doi:https://doi.org/10.1016/j.measurement.2022.111805.
  44. doi:https://doi.org/10.1016/j.jobe.2022.104098.
  45. doi:https://doi.org/10.1111/mice.12918.
  46. doi:https://doi.org/10.1007/s11263-021-01515-2.
  47. doi:https://doi.org/10.1016/j.autcon.2022.104299.
  48. doi:10.4121/6162a9b6-2a20-4600-8207-e9dcd53a264a.
  49. doi:https://doi.org/10.1109/5.726791.
  50. doi:https://doi.org/10.1145/3065386.
  51. doi:https://doi.org/10.48550/arXiv.1701.04128.
  52. doi:https://doi.org/10.1109/TITS.2016.2552248.
  53. doi:https://doi.org/10.48550/arXiv.1409.1556.
  54. doi:https://doi.org/10.1007/s11263-015-0816-y.
  55. doi:https://doi.org/10.48550/arXiv.1411.1792.
  56. doi:https://doi.org/10.1007/s00521-021-06279-x.
  57. doi:https://doi.org/10.48550/arXiv.1905.11946.
  58. doi:https://doi.org/10.1177/14759217211053776.
  59. doi:https://doi.org/10.1023/B:VISI.0000022288.19776.77.
  60. doi:https://doi.org/10.48550/arXiv.1907.02248.
  61. doi:https://doi.org/10.1007/s10851-018-0795-z.
  62. doi:https://doi.org/10.48550/arXiv.1711.05101.
  63. doi:https://doi.org/10.1090/S0033-569X-10-01172-0.
  64. doi:https://doi.org/10.1007/s10851-023-01170-x.
  65. doi:https://doi.org/10.1016/j.autcon.2022.104678.
  66. doi:https://doi.org/10.1109/TITS.2019.2910595.
  67. doi:https://doi.org/10.1016/j.neucom.2019.01.036.
  68. doi:https://doi.org/10.1111/mice.12412.
  69. doi:https://doi.org/10.1109/TITS.2015.2477675.
  70. doi:https://doi.org/10.1016/j.conbuildmat.2020.119397.
  71. doi:https://doi.org/10.1109/ACCESS.2018.2829347.
  72. doi:https://doi.org/10.3390/s20092557.
  73. doi:https://doi.org/10.1016/j.neucom.2022.01.051.
  74. doi:https://doi.org/10.1109/JAS.2023.123447.
  75. doi:https://doi.org/10.1016/j.autcon.2022.104436.
  76. doi:https://doi.org/10.1111/mice.12881.
  77. doi:https://doi.org/10.48550/arXiv.1802.02208.
  78. doi:https://doi.org/10.3390/coatings10020152.
  79. doi:https://doi.org/10.3390/s21092902.
  80. doi:https://doi.org/10.1016/j.engappai.2023.106142.
  81. doi:https://doi.org/10.1177/14759217221089571.
  82. doi:https://doi.org/10.13053/cys-23-2-3047.
Citations (1)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now