Detection of Moving Objects in Crowded Scenes using Trajectory Clustering via Conditional Random Fields Framework

Conditional Random Fields 구조에서 궤적군집화를 이용한 혼잡 영상의 이동 객체 검출

  • 김형기 ((주) 만도 ADAS 사업실 SYSTEM3팀) ;
  • 이광국 (한양대학교 전자컴퓨터통신공학과) ;
  • 김회율 (한양대학교 전자통신컴퓨터공학부)
  • Received : 2009.10.08
  • Accepted : 2010.05.11
  • Published : 2010.08.31

Abstract

This paper proposes a method of moving object detection in crowded scene using clustered trajectory. Unlike previous appearance based approaches, the proposed method employes motion information only to isolate moving objects. In the proposed method, feature points are extracted from input frames first and then feature tracking is followed to create feature trajectories. Based on an assumption that feature points originated from the same objects shows similar motion as the object moves, the proposed method detects moving objects by clustering trajectories of similar motions. For this purpose an energy function based on spatial proximity, motion coherence, and temporal continuity is defined to measure the similarity between two trajectories and the clustering is achieved by minimizing the energy function in CRFs (conditional random fields). Compared to previous methods, which are unable to separate falsely merged trajectories during the clustering process, the proposed method is able to rearrange the falsely merged trajectories during iteration because the clustering is solved my energy minimization in CRFs. Experiment results with three different crowded scenes show about 94% detection rate with 7% false alarm rate.

본 논문은 궤적을 군집화하여 혼잡한 영상에서 이동 객체를 검출하는 방법을 제안한다. 제안하는 방법은 객체의 외형 정보에 기반한 기존의 방법들과는 달리 객체의 움직임 정보만을 이용해 이동 객체를 검출한다. 이를 위하여 입력 영상의 매 프레임에서 특징점을 추출하며, 인접한 프레임간의 추적 과정을 통하여 특징점들의 궤적을 생성한다. 동일 객체에서 얻어진 궤적들은 유사한 움직임을 보일 것이라는 가정 하에 군집화 과정을 통하여 이동 객체를 검출한다. 궤적들의 군집화를 위하여 특징점 간의 위치, 움직임, 연속성에 기반한 에너지 함수로 궤적 간 유사도를 측정하였으며, conditional random fields (CRFs)를 이용하여 최적의 군집을 결정하였다. 기존의 궤적 군집화를 통한 이동 객체 검출 방법이 군집화 과정에서 한번 잘못 분류된 궤적은 잘못된 결과를 생성하는 것과는 달리, 제안한 방법에서는 군집화가 CRFs 상에서 에너지 최소화에 의해 수행되기 때문에 잘못 분류된 궤적이 반복 과정에서 다시 올바른 군집으로 재배열되는 것이 가능하다. 제안한 방법의 성능 측정을 위하여 서로 다른 혼잡도를 가지는 세 개의 영상을 이용하였으며, 약 94%의 검출률과 7%의 허위 경보율을 나타내었다.

Keywords

References

  1. S. J. McKenna, S. Jabri, Z. Duric, A. Rosenfeld, and H. Wechsler, "Tracking Groups of People," Computer Vision and Image Understanding, Vol.80, pp. 42-56, 2000. https://doi.org/10.1006/cviu.2000.0870
  2. S. A. Velastin, J. H. Yin, A. C. Davies, M. A. Vicencio-Silva, R. E. Allsop, and A. Penn, "Automated Measurement of Crowd Density and Motion using Image Processing," Seventh International Conference on Road Traffic Monitoring and Control, pp.127-132, April 1994.
  3. H. Celik, A. Hanjalic, and E. A. Hendriks, "Towards a Robust Solution to People Counting," Proc. of International Conference of Image Processing, pp.2401-2404, October 2006.
  4. C. R. Wren, A. Azarbayejani, T. Darrell, and A. Pentland, "Pfinder: Real-time Trakcking of the Human Body," IEEE Trans. on Pattern Analysis and Machine Intelligence, Vol.19, No. 7, pp. 780-785, 1997. https://doi.org/10.1109/34.598236
  5. G. Medioni, I. Cohen, F. Bremond, S. Hongeng, and R. Nevatia, "Event detection and analysis from video streams," IEEE Trans. on Pattern Analysis and Machine Intelligence, Vol.23, pp. 873-889, August 2001. https://doi.org/10.1109/34.946990
  6. G.-G. Lee, B. Kim, and W.-Y. Kim, "Automatic Estimation of Pedestrian Flow," First International Conference on Distributed Smart Cameras, pp. 291-296, September 2007.
  7. A. Utsumi and N. Tetsutani, "Human detection using geometrical pixel value structures," Fifth IEEE International Conference on Automatic Face and Gesture Recognition, pp.34-39, May 2002.
  8. T. Yang, J. Li, Q. Pan, C. Zhao, and Y. Zhu, "Active learning based pedestrian detection in real scenes," 18th International Conference on Pattern Recognition, Vol.4, pp.904-907, 2006.
  9. P. Viola, M.J. Jones, and D. Snow, "Detecting pedestrians using patterns of motion and appearance," Ninth IEEE International Conference on Computer Vision, Vol.2, pp.734-741, October 2003.
  10. G.J. Brostow and R. Cipolla, "Unsupervised bayesian detection of independent motion in crowds," IEEE Computer Society Conference on Computer Vision and Pattern Recognition, Vol.1, pp. 594-601, June 2006.
  11. V. Rabaud S. Belongie, "Counting crowded moving objects," IEEE Computer Society Conference on Computer Vision and Pattern Recognition, Vol.1, pp.705-711, June 2006.
  12. Y. Li and H. Ai, "Fast detection of independent motion in crowds guided by supervised learning," IEEE International Conference on Image Processing, Vol.3, pp.341-344, October 2007.
  13. E. Rosten and T. Drummond, "Fusing points and lines for high performance tracking," Tenth IEEE International Conference on Computer Vision, Vol. 2, pp. 17-21, October 2005.
  14. E. Rosten and T. Drummond, "Machine learning for high-speed comer detection," European Conference on Computer Vision, Vol.1, pp. 430-443, May 2006.
  15. C. Tomasi and T. Kanade, "Detection and tracking of point features," Carnegie Mellon University, CMU_CS_91-132, 1991.
  16. C. Tomasi and J. Shi, "Good features to track," IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pp. 593-600, June 1994.
  17. S. Birchfield, "Source code of the klt feature tracker," http://www.ces.clemson.edu/~stb/klt/, 2006.
  18. A. McCallum, J. Lafferty, and F. Pereira, "Conditional random fields: Probabilistic models for segmenting and labeling sequence data," Eighteenth International Conference on Machine Learning, pp.282-289, June 2001.
  19. 김형기, 이광국, 윤자영, 김재준, 김회율, "고정소수점 연산을 이용한 빠른 배경 모델링," 제 21회 영상처리 및 이해에 관한 워크샵 논문집(II), pp.345-348, 2009.
  20. P. Perez, "Markov random fields and images," Centrum voor Wiskunde en Informatica Quarterly, Vol. 11, pp. 413-437, 1998.
  21. Y. Yuan and C.T. Li, "Unsupervised clustering of gene expression time series with conditional random fields," Digital EcoSystems and Technologies Conference, pp. 571-576, Febrary 2007.