Marginalized Gromov Flow Particle Filter-Based Track-Before-Detect Algorithm for Marine Targets

Marine target tracking suffers from many difficulties such as weak echoes and complex background environments. In this paper, a marginalized Gromov particle flow particle filter-based track-before-detect (MPFPF-TBD) algorithm is proposed. Firstly, the invertible Gromov particle flow is utilized to incorporate the observation information, aiming to optimize proposal distribution. Secondly, the target state space is divided using a marginalization strategy, so the particle flow and extended Kalman methods can be utilized to calculate the nonlinear measurement-related state component and the conditionally linear state component, respectively. Finally, the K-distribution is applied to design the marine clutter model, and the track-before-detect method is utilized for energy accumulation. This algorithm is evaluated using simulated targets in simulated and measured sea clutter. The results indicate that, compared with the particle method, the LEDH-based particle flow method and the SFA-PF-TBD method, this algorithm exhibits advantages in estimation performance, computational efficiency and robustness.