CONDENSATION—Conditional Density Propagation forVisual Tracking

  • Authors:
  • Michael Isard;Andrew Blake

  • Affiliations:
  • Department of Engineering Science, University of Oxford, Oxford OX1 3PJ, UK;Department of Engineering Science, University of Oxford, Oxford OX1 3PJ, UK

  • Venue:
  • International Journal of Computer Vision
  • Year:
  • 1998

Quantified Score

Hi-index 0.00

Visualization

Abstract

The problem of tracking curves in dense visual clutter is challenging. Kalman filtering is inadequate because it is based on Gaussiandensities which, being unimo dal, cannot represent simultaneous alternativehypotheses. The Condensation algorithm uses“factored sampling”, previously applied to the interpretationof static images, in which the probability distribution of possibleinterpretations is represented by a randomly generated set. Condensation uses learned dynamical models, together with visualobservations, to propagate the random set over time. The result is highlyrobust tracking of agile motion. Notwithstanding the use of stochasticmethods, the algorithm runs in near real-time.