Development of Robust Crack Segmentation and Thickness Measurement Model Using Deep Learning 


Vol. 48,  No. 5, pp. 554-566, May  2023
10.7840/kics.2023.48.5.554


PDF
  Abstract

Investigation and measurement of crack is essential in maintaining safety of the facility. If crack is present, thorough examination of whether its development is in-progress is necessary, followed by precise measurement of its thickness. Due to the growing number of facilities requiring regular monitoring, automation of crack examination process is gaining attention. This paper introduces an deep learning model specializing on crack segmentation and its measurement based on image. For this task, Pohang Crack(POC), an original dataset was established : focused on properly reflecting the safety check environment, involving a sticker to measure the thickness. In addition, we propose the application of magic wand algorithm in the annotation of crack segmentation task for improving its credibility. DenseNet201-UNet crack segmentation model achieved the performance of 84.53 Crack IoU on the proposed dataset. Finally, through image skeletonization and euclidean distance transform on predicted segmentation mask, the thickest part of the crack was identified and measured pixel-wise which was then converted to cm measure in comparison with the pixel measure of the sticker. Through comparing the thickness of actual and predicted measurement, the margin of error was 0.09 cm, verifying its capability in the application on on-site safety 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]

SooMinLee, Gyeong-YeongKim, Dong-JuKim, "Development of Robust Crack Segmentation and Thickness Measurement Model Using Deep Learning," The Journal of Korean Institute of Communications and Information Sciences, vol. 48, no. 5, pp. 554-566, 2023. DOI: 10.7840/kics.2023.48.5.554.

[ACM Style]

SooMinLee, Gyeong-YeongKim, and Dong-JuKim. 2023. Development of Robust Crack Segmentation and Thickness Measurement Model Using Deep Learning. The Journal of Korean Institute of Communications and Information Sciences, 48, 5, (2023), 554-566. DOI: 10.7840/kics.2023.48.5.554.

[KICS Style]

SooMinLee, Gyeong-YeongKim, Dong-JuKim, "Development of Robust Crack Segmentation and Thickness Measurement Model Using Deep Learning," The Journal of Korean Institute of Communications and Information Sciences, vol. 48, no. 5, pp. 554-566, 5. 2023. (https://doi.org/10.7840/kics.2023.48.5.554)
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