Multi-Bernoulli filter based track-before-detect for Jump Markov models
Suqi Li, Wei Yi, Lingjiang Kong, Bailu Wang · 2014
This paper deals with the problem of simultaneously detecting and tracking multiple maneuvering targets. The multitarget, multi-Bernoulli (MeMber) filter based track-before-detect (TBD) is an attractive approach to detect and track targets at low signal-to-noise (SNR). However, MeMber-TBD with a fixed motion model is not general enough to accommodate maneuvering targets. In this paper, a new MeMber filter in the TBD context is proposed to cope with unknown and time-varying number of maneuvering targets. We extend the basic MeMber-TBD with Jump Markov System (JMS) multi-target models to accommodate target birth, death and switching dynamics. The recursive prediction and update equations of the proposed JMS-MeMber-TBD are derived and implemented using the sequential Monte Carlo (SMC) approximations. Simulation results for a challenging tracking scenario prove the effectiveness of the proposed algorithm.