Trajectory PHD and CPHD Filters With Unknown Detection Profile

Compared to the probability hypothesis density (PHD) and cardinalized PHD (CPHD) filters, trajectory (TPHD) CPHD (TCPHD) filters are for sets of trajectories, thus able produce estimates with better performance. In this paper, we develop TPHD TCPHD which can adaptively learn history unknown target detection probability, therefore they perform more robustly in scenarios where targets time-varying probabilities. These referred as (U-TPHD) (U-TCPHD) filters. By minimizing Kullback-Leibler divergence (KLD), U-TPHD U-TCPHD obtain, respectively, best Poisson independent identically distributed (IID) approximations over a set augmented trajectories. For computational efficiency, also propose that only consider profile at current time. Specifically, Beta-Gaussian mixture method is adopted implementations proposed BG-U-TPHD BG-U-TCPHD The $L$ -scan these much lower burden presented. Finally, various simulation results demonstrate achieve robust tracking performance adapt profile. Besides, it shows usually small value approximation has similar than large fraction cost.