DOI QR코드

DOI QR Code

Improvement of learning concrete crack detection model by weighted loss function

  • Received : 2020.08.12
  • Accepted : 2020.10.05
  • Published : 2020.10.30

Abstract

In this study, we propose an improvement method that can create U-Net model which detect fine concrete cracks by applying a weighted loss function. Because cracks in concrete are a factor that threatens safety, it is important to periodically check the condition and take prompt initial measures. However, currently, the visual inspection is mainly used in which the inspector directly inspects and evaluates with naked eyes. This has limitations not only in terms of accuracy, but also in terms of cost, time and safety. Accordingly, technologies using deep learning is being researched so that minute cracks generated in concrete structures can be detected quickly and accurately. As a result of attempting crack detection using U-Net in this study, it was confirmed that it could not detect minute cracks. Accordingly, as a result of verifying the performance of the model trained by applying the suggested weighted loss function, a highly reliable value (Accuracy) of 99% or higher and a harmonic average (F1_Score) of 89% to 92% was derived. The performance of the learning improvement plan was verified through the results of accurately and clearly detecting cracks.

본 연구에서는 가중치 오차 함수를 적용하여, 미세한 콘크리트 균열을 감지하는 U-Net 모델을 만들 수 있도록 개선 방안을 제안한다. 콘크리트 균열은 안전을 위협하는 요소이기 때문에 그 상태를 주기적으로 파악하고 신속하게 초기 대응을 하는 것이 중요하다. 하지만 현재는 점검자가 직접 육안으로 검사하고 평가하는 외관 검사법이 주로 사용되고 있다. 이는 정확성뿐만 아니라 비용과 시간, 안전성 측면에서도 한계점을 가진다. 이에 콘크리트 구조물에 생성되는 미세한 균열을 신속하고 정밀하게 탐지할 수 있도록 딥러닝을 활용한 기술들이 연구되고 있다. 본 연구에서 U-Net을 활용한 균열 탐지를 시도한 결과, 미세한 균열을 탐지하지 못하는 것을 확인하였다. 이에 제시한 가중치 오차 함수를 적용하여 학습한 모델에 대해 성능을 검증한 결과, 정확도(Accuracy) 99% 이상, 조화평균(F1_Score) 89%에서 92%의 신뢰성 높은 수치를 도출해내었고, 미세한 균열을 정확하고 선명하게 탐지한 결과를 통해 학습 개선 방안의 성능을 검증하였다.

Keywords

References

  1. Ministry of Land, Infrastructure and Transport, Detailed guidelines for safety and maintenance of facilities, 2019
  2. UNIST(ULSAN NATIONAL INSTITUTE OF SCIENCE AND TECHNOLOGY), Apparatus and Method for Detecting Construction Crack, 1020170100990, 2016.02.26., 2017.10.19.
  3. Junhee Kim, Yoon-Soo Shin and Kyung-Won Min, "Line Laser Image Processing for Automated Crack Detection of Concrete Structures", Journal of the Computational Structural Engineering Institute of Korea, Vol. 31, No. 3, pp.147-153,June 2018. DOI: 10.7734/COSEIK.2018.31.3.147
  4. Jung, Seo-Young, Yu, Jung-Ho, "A Proposal of Automatic Condition Evalutation Process based on Unsupervised Learning", Proceedings of KICEM Annual Conference, pp.79-80, Korea, Nov. 2019
  5. Jieun Lee, Hee-Sun Kim, Nayoung Kim, Eun-Mi Ryu, and Je-Won Kang, "Learning to Detect Cracks on Damaged Concrete Surfaces Using Two-Branched Convolutional Neural Network", SENSORS vol. 19, no. 21, n. pag., November 2019. DOI: 10.3390/s19214796
  6. Cho, Soojin, Kim, Byunghyun, "Automated Concrete Crack Detection Technique Using Deep Learning", Journal of Structural Integrity and Maintenance, Vol 23, No. 3, pp. 13-21, 2019
  7. Exxact Croporation, Atrous Convolutions & U-Net Architectures for Deep Learning: A Brief History, https://blog.exxactcorp.com/atrous-convolutions-u-net-architectures-for-deep-learning-a
  8. Kim, Ah-Ram, Kim, Donghyeon, Byun, Yo-Seph, Lee, Seong-Won, "Crack Detection of Concrete Structure Using Deep Learning and Image Processing Method in Geotechnical Engineering", JOURNAL OF THE KOREAN GEOTECHNICAL SOCIETY, pp.145-154, Korea, Dec. 2018, DOI: 10.7843/kgs.2018.34.12.145
  9. Seo, Wonju, UNET, https://wewinserv.tistory.com/30
  10. Kang, Eunsuk, U-NET, https://modulabs-biomedical.github.io/U_Net
  11. Olaf Ronneberger, Philipp Fischer, Thomas Brox, "U-Net: Convolutional Networks for Biomedical Image Segmentation", MICCAI, May. 2015, DOI: 10.1007/978-3-319-24574-4_28
  12. Kim, Chungyun, Evaluation Method of Classification Model, https://velog.io/@skyepodium/%EB%B6%84%EB%A5%98-%EB%AA%A8%EB%8D%B8-%ED%8F%89%EA%B0%80-%EB%B0%A9%EB%B2%95
  13. Caglar Firat Ozgenel, Concrete Crack Segmentation Dataset, https://data.mendeley.com/datasets/jwsn7tfbrp/1
  14. Um, Sungbum, Artificial Neural Network(ANN) #4-1 Learning(Loss function, Error function), https://umbum.dev/210