A VISION-BASED TRACKING SYSTEM FOR MONITORING TRAFFIC ACTIVITIES AT INTERSECTIONS

Monitoring activities in outdoor traffic scenes has applications in areas such as data collection, monitoring for traffic congestion, and surveillance for safety. This project develops a camera based sensing system for monitoring activities at intersections. Two different components of the system are the vision module and the incident detection module, the former being responsible for providing information about the scene such as vehicle position, velocity, shape, etc. The incident detection module is responsible for detecting incidents such as collisions and near misses. The focus of this work is on the vision module, in particular vehicle tracking. Tracking in outdoor scenes using camera is affected by illumination changes, traffic congestion, clutter, and stop-and-go motion of the vehicles. In this paper, it is shown that decent tracking, despite these conditions by using two different tracking modalities can be attained. The first one is a region tracker, which uses the results of the scene segmentation for tracking. The second tracker known as the Mean Shift Tracker, makes use of the color distribution of the vehicles. The results from the two trackers, namely, the vehicle's position are combined sequentially in each step.

Language

  • English

Media Info

  • Features: Figures; References;
  • Pagination: 12p

Subject/Index Terms

Filing Info

  • Accession Number: 00976134
  • Record Type: Publication
  • Files: TRIS
  • Created Date: Jul 9 2004 12:00AM