Moving Target Classification and Tracking from Real-time Video

  • Authors:
  • Alan J. Lipton;Hironobu Fujiyoshi;Raju S. Patil

  • Affiliations:
  • -;-;-

  • Venue:
  • WACV '98 Proceedings of the 4th IEEE Workshop on Applications of Computer Vision (WACV'98)
  • Year:
  • 1998

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper describes an end-to-end method for extractingmoving targets from a real-time video stream, classifyingthem into predefined categories according to image-basedproperties, and then robustly tracking them. Movingtargets are detected using the pixel wise difference betweenconsecutive image frames. A classification metric is appliedthese targets with a temporal consistency constraintto classify them into three categories: human, vehicle orbackground clutter. Once classified, targets are tracked bya combinationof temporal differencing and template matching.The resulting system robustly identifies targets of interest,rejects background clutter, and continually tracks overlarge distances and periods of time despite occlusions, appearancechanges and cessation of target motion.