Tracking Across Multiple Cameras With Disjoint Views

  • Authors:
  • Omar Javed;Zeeshan Rasheed;Khurram Shafique;Mubarak Shah

  • Affiliations:
  • -;-;-;-

  • Venue:
  • ICCV '03 Proceedings of the Ninth IEEE International Conference on Computer Vision - Volume 2
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

Conventional tracking approaches assume proximity inspace, time and appearance of objects in successive observations.However, observations of objects are often widelyseparated in time and space when viewed from multiplenon-overlapping cameras. To address this problem, wepresent a novel approach for establishing object correspondenceacross non-overlapping cameras. Our multi-cameratracking algorithm exploits the redundance in paths thatpeople and cars tend to follow, e.g. roads, walk-ways orcorridors, by using motion trends and appearance of objects,to establish correspondence. Our system does notrequire any inter-camera calibration, instead the systemlearns the camera topology and path probabilities of objectsusing Parzen windows, during a training phase. Oncethe training is complete, correspondences are assigned usingthe maximum a posteriori (MAP) estimation framework.The learned parameters are updated with changing trajectorypatterns. Experiments with real world videos are reported,which validate the proposed approach.