Domain Adaptive Deep Learning-Based Crack Detection for Building Inspection 


Vol. 48,  No. 5, pp. 567-580, May  2023
10.7840/kics.2023.48.5.567


PDF
  Abstract

In building safety inspection, the inspection of cracks plays a significant role in the process since it intuitively indicates the dangerousness of the facility. To aid the automation of crack inspection, development of an artificial intelligence-based model has been ongoing. However, the development of a robust crack detection model faltered due to the limitations of existing crack datasets. Depending on the type and location of the crack on the wall, cracks vary greatly requiring a wide variety of domains in the training dataset. In this paper, we propose a domain-general crack detection model specializing in building safety inspection environments along with introducing POC(Pohang Crack), a crack dataset consisting of 11,466 images manually collected and annotated, ensuring diversity of characteristics and distribution of data through stratified sampling method. By comparing the detection performance of deep learning-based YOLO models, YOLO-Cr was developed, showing the highest mAP(mean Average Precision) of 0.915. Moreover, we introduce a detection system utilizing real-time scanning and UAV to aid the visual inspection. It is expected that this study will contribute to the automation of safety inspection of various structures and increase the objectivity and efficiency of inspection.

  Statistics
Cumulative Counts from November, 2022
Multiple requests among the same browser session are counted as one view. If you mouse over a chart, the values of data points will be shown.


  Related Articles
  Cite this article

[IEEE Style]

Ji-HoKim, Gyeong-YeongKim, Dong-JuKim, "Domain Adaptive Deep Learning-Based Crack Detection for Building Inspection," The Journal of Korean Institute of Communications and Information Sciences, vol. 48, no. 5, pp. 567-580, 2023. DOI: 10.7840/kics.2023.48.5.567.

[ACM Style]

Ji-HoKim, Gyeong-YeongKim, and Dong-JuKim. 2023. Domain Adaptive Deep Learning-Based Crack Detection for Building Inspection. The Journal of Korean Institute of Communications and Information Sciences, 48, 5, (2023), 567-580. DOI: 10.7840/kics.2023.48.5.567.

[KICS Style]

Ji-HoKim, Gyeong-YeongKim, Dong-JuKim, "Domain Adaptive Deep Learning-Based Crack Detection for Building Inspection," The Journal of Korean Institute of Communications and Information Sciences, vol. 48, no. 5, pp. 567-580, 5. 2023. (https://doi.org/10.7840/kics.2023.48.5.567)
Vol. 48, No. 5 Index

Search
(IN TITLE, AUTHOR, ABSTRACT,KEYWORDS)

Advanced Search ▼

POPULAR KEYWORDS
(TOP 10 KEYWORDS)

Recent Publications
(LAST 3 YEARS)






pISSN : 1226 - 4717
eISSN : 2287 - 3880