On models of observing and tracking ground targets based on Hidden Markov Processes and Bayesian networks


  • Kaijser Thomas

Publish date: 2004-01-01

Report number: FOI-R--1192--SE

Pages: 53

Written in: Swedish


Hidden Markov Processes and Bayesian nets are areas within probability theory which have found many practical applications within the last decade. In this report a general description is given of how the tracking of ground targets can be modelled as Hidden Markov processes and Bayesian nets. An important property of Bayesian nets is that new information can be handled rapidly and spread quickly within the net. In the report some convergence results for Hidden Markov processes are presented and proven, results which lead to the following simple consensus principle: "If two persons are watching a process through an observation system and they agree upon the observation probabilities for every possible state of the process, then their opinions of the state of the process will converge, as more and more observations are obtained, irrespectively of how different their initial opinions of the state of the process may have been."