DOI QR코드

DOI QR Code

A Study on Deep Learning-based Pedestrian Detection and Alarm System

딥러닝 기반의 보행자 탐지 및 경보 시스템 연구

  • Kim, Jeong-Hwan (Dept. of Computer Science & Eng., Seoul National Univ. of Science & Technology) ;
  • Shin, Yong-Hyeon (Dept. of Computer Science & Eng., Seoul National Univ. of Science & Technology)
  • 김정환 (서울과학기술대 컴퓨터공학과) ;
  • 신용현 (서울과학기술대 컴퓨터공학과)
  • Received : 2019.04.10
  • Accepted : 2019.08.12
  • Published : 2019.08.31

Abstract

In the case of a pedestrian traffic accident, it has a large-scale danger directly connected by a fatal accident at the time of the accident. The domestic ITS is not used for intelligent risk classification because it is used only for collecting traffic information despite of the construction of good quality traffic infrastructure. The CNN based pedestrian detection classification model, which is a major component of the proposed system, is implemented on an embedded system assuming that it is installed and operated in a restricted environment. A new model was created by improving YOLO's artificial neural network, and the real-time detection speed result of average accuracy 86.29% and 21.1 fps was shown with 20,000 iterative learning. And we constructed a protocol interworking scenario and implementation of a system that can connect with the ITS. If a pedestrian accident prevention system connected with ITS will be implemented through this study, it will help to reduce the cost of constructing a new infrastructure and reduce the incidence of traffic accidents for pedestrians, and we can also reduce the cost for system monitoring.

보행자 교통사고의 경우 사고 발생 시 사망사고로 연결되는 위험성이 있다. 국내 지능형 교통시스템(ITS)은 질 좋은 교통 인프라를 구축하고 있음에도 불구하고, 거의 교통정보 수집에만 이용되고 있어, 위험상황 발생 시 지능적인 위험 요소 분류가 이루어지지 않고 있다. 본 연구에서 제안하는 시스템의 주요 구성 요소인 CNN 기반의 보행자 탐지 분류 모델의 경우 제한적인 환경에서 설치 운영되는 것을 가정하여 임베디드 시스템 기반으로 구현되었다. 기존 YOLO의 인공신경망 모델을 개선하여 My-Tiny-Model3라는 새로운 모델을 생성하였고, 20,000번의 반복 학습 기준으로 평균 정확도 86.29%와 21.1 fps의 실시간 탐지 속도 결과를 보였다. 그리고, 이러한 탐지 시스템을 기반으로 하여 ITS 체계와 연계 가능한 시스템 구현 및 프로토콜 연동 시나리오를 구성하였다. 본 연구를 통해 기존 ITS 체계와 연동하는 보행자 사고 방지 시스템을 구현한다면, 새로운 인프라 구축비용을 절감하고 보행자 교통사고 발생률을 줄이는 데 도움이 될 것이다. 또한, 기존의 시스템 감시인력 소요에 따른 비용 또한 줄일 수 있을 것으로 기대된다.

Keywords

References

  1. Dalal N. and Triggs B.(2005), "Histograms of Oriented Gradients for Human Detection," IEEE Computer Society, pp.886-893.
  2. Darknet J. Redmon.(2016), Open source neural networks inc, http://pjreddie.com/darknet/2018.08.20.
  3. Everingham M., Christopher L. V. G., Williams K. I., Winn J. and Zisserman A.(2010), "The PASCAL Visual Object Classes (VOC) Challenge," International Journal of Computer Vision manuscript, p.11.
  4. https://github.com/AlexeyAB/Yolo_mark 2018.09.24.
  5. Kim J., Lee D. and Lee M.(2016), "Lane Detection System using CNN," Institute Embedded Engineering of Korea, pp.163-169.
  6. Kim Y. M., Lee I., Yoon I., Han T. and Kim C.(2018), "CCTV Object Detection with Background Subtraction and Convolutional Neural Network," KIISE Transactions on Computing Practices, vol. 24, no. 3, pp.151-156. https://doi.org/10.5626/KTCP.2018.24.3.151
  7. Krizhevsky A., Sutskever I., Geoffrey and Hinton E.(2012), "ImageNet Classification with Deep Convolutional Neural Networks," NIPS'12 Proceedings of the 25th International Conference on Neural Information Processing Systems, vol. 1, pp.1097-1105.
  8. Lee Y. and Shin J.(2018), "DNN Based Multi-spectrum Pedestrian Detection Method Using Color and Thermal Image," Journal of Broadcast Engineering, vol. 23, no. 3, pp.361-368. https://doi.org/10.5909/JBE.2018.23.3.361
  9. Redmon J. and Farhadi A.(2017), "Yolo9000: Better, faster, stronger. In Computer Vision and Pattern Recognition," IEEE Conference on, pp.6517-6525.
  10. Redmon J. and Farhadi A.(2018), YOLOv3: An Incremental Improvement, arXiv, pp.1-4.
  11. Rye J. D., Park S., Park S., Kwon C. and Yun I.(2018), "A Study for Development of Expressway Traffic Accident Prediction Model Using Deep Learning," The Journal of The Korea Institute of Intelligent Transport Systems, vol. 17, no. 4, pp.14-25.
  12. TAAS(Cyber National Police Agency, Road Traffic Authority) statistics for 2017.
  13. The standard of UTIS(Urban Traffic Information Systems)(2014), PART II. Application Layer Protocol, pp.27-55.