Multiple transition mode multiple target track-before-detect with partitioned sampling

In this paper, we extend the multiple model track-before-detect method to track all possible target combinations at low signal-to-noise ratios. Given a maximum number of targets, the method estimates the posterior probability density function of the multitarget state vector, the corresponding target existence probabilities, and the probabilities of all possible target combinations. As the particle filter implementation of this method requires a large number of particles to achieve high tracking performance, we propose an efficient partition based proposal function method by partitioning the multiple target space into a set of single target spaces. We also integrate the Markov chain Monte Carlo Metropolis-Hastings method into the particle proposal process to improve sample diversity. The proposed algorithm is validated by tracking five targets in very low signal-to-noise ratios (SNRs).