대한토목학회논문집 (KSCE Journal of Civil and Environmental Engineering Research)

  • 제43권1호
  • /
  • Pages.99-107
  • /
  • 2023
  • /
  • 1015-6348(pISSN)
  • /
  • 2799-9629(eISSN)
대한토목학회 (Korean Society of Civil Engeneers)
권호 표지
DOI QR코드

DOI QR Code

인공지능 학습용 토공 건설장비 영상 데이터셋 구축 및 타당성 검토

Building-up and Feasibility Study of Image Dataset of Field Construction Equipments for AI Training

  • 나종호 (한국건설기술연구원) ;
  • 신휴성 (한국건설기술연구원 미래스마트건설연구본부) ;
  • 이재강 (한국건설기술연구원 미래스마트건설연구본부) ;
  • 윤일동 (한국외국어대학교 컴퓨터공학과)
  • Na, Jong Ho (Korea Institute of Construction Technology, Hankuk University of Foreign Studies) ;
  • Shin, Hyu Soun (Korea Institute of Construction Technology) ;
  • Lee, Jae Kang (Korea Institute of Construction Technology) ;
  • Yun, Il Dong (Hankuk University of Foreign Studies)
  • 투고 : 2022.08.30
  • 심사 : 2022.11.22
  • 발행 : 2023.02.01
PDF 다운로드
⟨ 이전 논문 다음 논문 ⟩

초록

최근 건설 현장의 안전사고 비율은 전체 산업에서 가장 높은 비중을 차지한다. 인공지능 기술을 건설 현장에 접목하기 위해서는 기초 학습 자료로 활용될 수 있는 데이터셋 확보가 필수적이다. 본 논문에서는 실제 현장 확보를 통해 원천 데이터를 수집하였으며, 토목 현장에서 주로 운용되고 있는 주요 건설장비 객체를 선정하고 약 9만장의 정지영상 데이터셋 가공을 통해 최적의 학습 데이터셋 구축을 완료하였다. 또한, 객체 인식분야의 대표적인 모델인 YOLO를 활용하여 구축된 데이터의 검증 작업을 수행하였고 90 % 근접한 검출 성능을 확인해 데이터 신뢰성을 확보하였다. 본 연구에서 사용되는 학습 데이터셋은 공공데이터포털에서 활용 가능하도록 공개를 완료하였다. 본 데이터셋은 향후 건설안전 분야의 객체 인식 기술의 건설현장 적용을 위한 기반 데이터로 활용 가능하리라 판단된다.

Recently, the rate of death and safety accidents at construction sites is the highest among all kinds of industries. In order to apply artificial intelligence technology to construction sites, it is essential to secure a dataset which can be used as a basic training data. In this paper, a number of image data were collected through actual construction site, for which major construction equipment objects mainly operated in civil engineering sites were defined. The optimal training dataset construction was completed by annotation process of about 90,000 image dataset. Reliability of the dataset was verified with the mAP of over 90 % in use of YOLO, a representative model in the field of object detection. The construction equipment training dataset built in this study has been released which is currently available on the public data portal of the Ministry of Public Administration and Security. This dataset is expected to be freely used for any application of object detection technology on construction sites especially in the field of construction safety in the future.

키워드

과제정보

본 논문은 한국건설기술연구원 주요사업으로 지원을 받아 수행된 연구(버츄얼 컨테이너 기반 독립적 소규모 건설현장 안전관리플랫폼 구축)로 이에 감사드립니다.

참고문헌

  1. Everingham, M., Van Gool, L., Williams, C. K., Winn, J. and Zisserman, A. (2010). "The pascal visual object classes (VOC) challenge." International Journal of Computer Vision, Vol. 88, No. 2, pp. 303-338. https://doi.org/10.1007/s11263-009-0275-4
  2. Girshick, R. (2015). "Fast r-cnn." In Proceedings of the IEEE International Conference on Computer Vision, Santiago, Chile, pp. 1440-1448.
  3. Girshick, R., Donahue, J., Darrell, T. and Malik, J. (2015). "Regionbased convolutional networks for accurate object detection and segmentation." IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 38, No. 1, pp. 142-158.
  4. Korea Institute of Civil Engineering and Bulding Technology (KICT) (2021). Public data portal site, Available at: https://www.data.go.kr/data/15095287/fileData.do (Accessed: November 22, 2022).
  5. Korea Meteorological Administration (KMA) (2021). Weather information, Available at: https://www. kma.go.kr/eng/index.jsp (Accessed: December 28, 2021).
  6. Ministry of Employment and Labor (MOEL) (2020). Industrial accident status (in Korean).
  7. Ministry of Government Legislation (MGL) (2022). Construction machinery management act (in Korean).
  8. Ministry of Land, Infrastructure and Transport (MOLIT) (2014). Development of risk factor for construction project, MOLIT Research Report (in Korean)
  9. Powers, D. M. (2010). "Evaluation: From precision, recall and F-measure to ROC, informedness, markedness and correlation." arXiv preprint, arXiv:2010.16061.
  10. Redmon, J. and Farhadi, A. (2017). "YOLO9000: Better, faster, stronger." In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, HI, USA, pp. 7263-7271.
  11. Redmon, J. and Farhadi, A. (2018). "Yolov3: An incremental improvement." In Computer Vision and Pattern Recognition, Vol. 1804, pp. 1-6.
  12. Ren, S., He, K., Girshick, R. and Sun, J. (2015). "Faster R-CNN: towards real-time object detection withregion proposal networks." Proceedings of the Advances in Neural Information Processing Systems, pp. 91-99.
  13. Seoul Metropolitan Government (2021). 10 Safety measures for small and medium-sized construction sites in Seoul, Seoul Metropolitan Government, Local Construction Safety Center (in Korean).
  14. Wada, K. (2018). Labelme, Github repository, Available at: https://github.com/wkentaro/labelme (Accessed: November 27, 2022).
  15. Xiao, B. and Kang, S. C. (2021). "Development of an image data set of construction machines for deep learning object detection." Journal of Computing in Civil Engineering, Vol. 35, No. 2.