The Journal of the Convergence on Culture Technology (문화기술의 융합)

The International Promotion Agency of Culture Technology (국제문화기술진흥원)
권호 표지
DOI QR코드

DOI QR Code

Accident Detection System for Construction Sites Using Multiple Cameras and Object Detection

다중 카메라와 객체 탐지를 활용한 건설 현장 사고 감지 시스템

  • 김민형 (한국기술교육대학교 컴퓨터공학과) ;
  • 감민성 (한국기술교육대학교 컴퓨터공학부) ;
  • 류호성 (한국기술교육대학교 컴퓨터공학부) ;
  • 박준혁 (한국기술교육대학교 컴퓨터공학부) ;
  • 전민수 (한국기술교육대학교 컴퓨터공학부) ;
  • 최형우 (한국기술교육대학교 컴퓨터공학부) ;
  • 민준기 (한국기술교육대학교 컴퓨터공학부)
  • Received : 2023.07.30
  • Accepted : 2023.09.05
  • Published : 2023.09.30
Download PDF
⟨ Previous Next ⟩

Abstract

Accidents at construction sites have a very high rate of fatalities due to the nature of being prone to severe injury patients. In order to reduce the mortality rate of severely injury patients, quick response is required, and some systems that detect accidents using AI technology and cameras have been devised to respond quickly to accidents. However, since existing accident detection systems use only a single camera, there are blind spots, Thus, they cannot detect all accidents at a construction site. Therefore, in this paper, we present the system that minimizes the detection blind spot by using multiple cameras. Our implemented system extracts feature points from the images of multiple cameras with the YOLO-pose library, and inputs the extracted feature points to a Long Short Term Memory-based recurrent neural network in order to detect accidents. In our experimental result, we confirme that the proposed system shows high accuracy while minimizing detection blind spots by using multiple cameras.

건설 현장의 사고는 중증외상환자가 발생하기 쉬운 특성 탓에 사망으로 이어지는 비율이 매우 높다. 중증외상환자의 사망률을 줄이기 위해서는 빠른 대처가 필요하며, 빠른 사고 대처를 위해 인공지능 기술과 카메라를 이용하여 사고를 감지하는 시스템들이 개발되었다. 그러나 기존 사고 감지 시스템들은 단일 카메라만을 사용하여, 사각지대로 인해 건설 현장의 모든 사고를 감지하기에 한계가 있다. 따라서, 본 논문에서는 다수의 카메라를 사용하여 감지 사각지대를 최소화하는 시스템을 구현하였다. 구현된 시스템은 다수의 카메라의 영상에서 YOLO-pose 라이브러리로 특징점을 추출하고, 추출된 특징점을 장단기 메모리(Long Short Term Memory) 기반 순환신경망에 입력하여 사고를 감지하였다. 실험 결과, 우리는 제안하는 시스템이 복수의 카메라 사용으로 감지 사각지대를 최소화하면서도 높은 정확도를 가지는 것을 확인하였다.

Keywords

Acknowledgement

이 논문은 2023년도 한국기술교육대학교 교수 교육연구진흥과제 지원에 의하여 연구되었음.

References

  1. Ministry of Employment and Labor "Additional Statistics on Industrial Accidents in 2022" 2022. https://www.moel.go.kr/policy/policydata/view.do?bbs_seq=20230100992
  2. J.C. Yang and J.D. Moon, "The effects of prehospital care on on-scene time in patients with major trauma" The Korean Journal of Emergency Medical Services, Vol. 24, No. 1, pp. 67-76, 2020. doi: 10.14408/KJEMS.2020.24.1.067n
  3. S. Jeong, S. Kang, and I. Chun, "Human-skeleton based fall-detection method using LSTM for manufacturing industries" 2019 34th International Technical Conference on Circuits/Systems, Computers and Communications (ITC-CSCC), pp. 1-4, 2019 doi: 10.1109/ITC-CSCC.2019.8793342
  4. J. Redmon, S. Divvala, R. Grishick and A. Farhadi, "You only look once: Unified, real-time object detection" Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 779-788, 2016. doi: 10.1109/CVPR.2016.91
  5. A. K. Jain, J. Mao and K. M. Mohiuddin,"Artificial neural networks: A tutorial" Computer, Vol. 29, No. 3, pp. 31-44, 1996. doi: 10.1109/2.485891
  6. B.S. Hwang, J.H. Kim, Y.R. Lee, C.U. Kyeong, J.H. Seon, Y.G. Sun, J.Y. Kim. "Performance of Exercise Posture Correction System Based on Deep Learning" The Journal of The Institute of Internet, Broadcasting and Communication, Vol. 22, No. 5, pp. 177-183, 2022 doi: doi.org/10.7236/JIIBC.2022.22.5.177
  7. I.S. Oh, "Machine Learning" Hanbit Academy, 2017.
  8. S. Alex, "Fundamentals of recurrent neural network (RNN) and long short-term memory (LSTM) network" Physica D: Nonlinear Phenomena, Vol. 404, pp. 132306, 2020. doi: 10.1016/j.physd.2019.132306
  9. S. Hochreiter, "The vanishing gradient problem during learning recurrent neural nets and problem solutions" International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems, Vol. 6, No. 3, pp. 107-116, 1998. doi: 10.1142/S0218488598000094
  10. J. Lu, M. Nguyen and W.Q. Yan, "Deep learning methods for human behavior recognition" 2020 35th International Conference on Image and Vision Computing New Zealand (IVCNZ). IEEE, pp 1-6, 2020. doi: 10.1109/IVCNZ51579.2020.9290640
  11. Y. Yoon and T. Oh, "A Study on the Improvement of Construction Site Worker Detection Performance Using YOLOv5 and Open-Pose" The journal of the convergence on culture technology, Vol. 8, No. 5, pp. 735-740, 2022. doi: 10.17703/JCCT.2022.8.5.735.
  12. Y.W. Lee, J.H. Park, S.Y. Sin, "Implementation of Fall Detection Based on CNN-LSTM" Vol. 4 7, No. 2, pp. 340-347, 2022. doi:10.7840/kics.2022.47.2.340
  13. W.H. Choi, C.D. Kwon, B.S. Yoo, M.H. Kim and J.K. Min. "Accident Detection System Based on RNN Exploiting Keypoints and LSTM" KIISE Transactions on Computing Practices, Vol. 29, No. 7, pp. 309-315, 2023. doi: 10.5626/KTCP.2023.29.7.309
  14. R. Girshick, J. Donahue, T. Darre and J. Malik. "Rich feature hierarchies for accurate object detection and semantic segmentation" Proceedings of the IEEE conference on computer vision and pat tern recognition, pp. 580-587, 2014. doi: 10.1109/CVPR.2014.81
  15. N.J. Kwak, D.J. Kim. "A study on Detecting the Safety helmet wearing using YOLOv5-S model and transfer learning" International Journal of Advanced Culture Technology, Vol.10, No.1, pp. 30 2-309, 2022 doi: 10.17703/IJACT.2022.10.1.302
  16. D. Maji, S. Nagori, M. Mathew and D. Poddar. "Yolo-pose: Enhancing yolo for multi person pose estimation using object keypoint similarity loss" Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 2637-2646, 2022. doi: 10.1109/CVPRW56347.2022.00297
  17. N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever and R. Salakhutdinov. "Dropout: a simple way to prevent neural networks from overfitting" The journal of machine learning research, Vol. 15, No. 1, pp. 1929-1958, 2014.
  18. AIHub, "Senior Abnormal Behavior Video" 2022. https://aihub.or.kr/
  19. N.V. Chawla, K.W. Bowyer, L.O. Hall and W.P. Kegelmeyer. "SMOTE: synthetic minority over-sampling technique" Journal of artificial intelligence research, Vol. 16, pp. 321-357, 2002. doi: 10.48550/arXiv.1106.1813