MULTIPLE TARGET TRACKING USING HIDDEN MARKOV-MODELS

Abstract

The application of hidden Markov models (HMM) to the problem of tracking multiple targets is discussed. The tracker generates multiple discrete Viterbi tracks and automatically accounts for track iteration, termination, and ambiguous measurements. The tracker is not smoothing-based, as are most existing systems such as Kalman and PDA (probabilistic data association) trackers, but is discrete in the sense of the finite state Viterbi algorithm. Simulation shows that in some cases it is possible to avoid the route of data association and directly compute the maximum likelihood mixed track.