Pedestrian Tracking by On-Board Wideview Monocular Camera

This research presents a general algorithm for pedestrian tracking by on-board monocular camera which can be applied to cameras of various view ranges in unified manner. The Spatio-Temporal MRF(S-T MRF) model extracts and tracks foreground objects, under the assumption that background motion of camera can be approximated by a linear function. By this method, both pedestrians and non-pedestrians should be segmented from background motion. Then, Region of Interest (ROI) is classified by a cascaded classifier. However, segmentation result sometime contains error due to the motion of pedestrian. The authors improved the performance of the system by contaminating errors including in foreground objects. After considering the geometrical understanding, the motion difference, and HOG information, the authors combined those features to correct the result of segmentation. The authors confirmed this combination approach drastically improved the performance of pedestrian detection.

  • Availability:
  • Supplemental Notes:
    • Abstract reprinted with permission from Intelligent Transportation Society of America.
  • Corporate Authors:

    ITS America

    1100 17th Street, NW, 12th Floor
    Washington, DC  United States  20036
  • Authors:
    • Shibayama, Yuuki
    • Kim, HyungKwan
    • Kamijo, Shunsuke
  • Conference:
  • Publication Date: 2011

Language

  • English

Media Info

  • Media Type: Digital/other
  • Features: Figures; Photos; References; Tables;
  • Pagination: 14p
  • Monograph Title: 18th ITS World Congress, Orlando, 2011. Proceedings

Subject/Index Terms

Filing Info

  • Accession Number: 01447790
  • Record Type: Publication
  • Files: TRIS
  • Created Date: Sep 28 2012 4:58PM