Optimal and suboptimal shape tracking based on multiple switched dynamic models

摘要

Object tracking based on multiple models has been recently advocated as a way to tackle sudden changes of shape or motion parameters. This paper addresses the estimation of time-varying parameters described by a bank of shared state stochastic models, switched according to a probabilistic mechanism (hidden Markov chain). A state estimation algorithm is proposed, based on the propagation of Gaussian mixtures in a multi-model framework. For preventing mode explosion a pruning strategy combining mode elimination and merging is used. This is shown to be better than employing either just elimination or merging. Examples dealing with image processing of moving objects are provided.