Particle Filtering Track-Before-Detect for Maneuvering Target in Jump Markov Systems

Maneuvering target is assumed to pop up and disappear according to the Jump Markov System (JMS) in clutters. A new particle filter based track-before-detect algorithm with unscented transform (UT) and interacting scheme is proposed in this paper. The method or particle splitting and direct sampling arc addressed in the previous work, which are disadvantageous to a particle filtering algorithm. The UT transform and interacting scheme in the algorithm avoids both the particle splitting and direct sampling and eases the problem of particle degeneracy. Through the proposed algorithm the target is tracked before the detection, by jointly estimating the target state and detecting the presence of target, eventually to its track. The whole frame of the algorithm is well structured in this paper. Simulations show that our approach saves the combinatorial complexity in mode-matched filtering. The paper deals with only a single target in clutters.