Standard Article

Statistical Methods for Target Tracking

  1. Simon Maskell

Published Online: 16 MAR 2009

DOI: 10.1002/9780470050118.ecse645

Wiley Encyclopedia of Computer Science and Engineering

Wiley Encyclopedia of Computer Science and Engineering

How to Cite

Maskell, S. 2009. Statistical Methods for Target Tracking. Wiley Encyclopedia of Computer Science and Engineering. 2820–2829.

Author Information

  1. Qineti Q, Worcester, United Kingdom

Publication History

  1. Published Online: 16 MAR 2009

Abstract

Tracking is the algorithmic process of updating of a perception of the time-evolving state of the stochastic world as a noisy stream of data is received. To do this, one needs to define statistical models for the evolution of the state and the relationship between the state and the data. The definition of the state and the stream are problem specific. Indeed, these definitions define the problem that a specific tracking algorithm is tackling. An algorithmic implementation of an exact solution to this mathematical problem can be defined in cases when the models have certain mathematical properties. An example is the Kalman filter. However, often the models require approximation: This article reviews both exact and approximate solutions based on the Kalman filter and particle filter. It then goes on to describe the use of multihypothesis techniques and multitarget tracking algorithms. Finally, the mean-shift tracking paradigm, which is popular in the computer vision literature, is discussed.

Keywords:

  • Kalman filter;
  • particle filter;
  • tracking;
  • multiple hypothesis;
  • multitarget;
  • mean-shift;
  • video;
  • radar