Adaptive continuous-discrete trajectory PHD filter with out-of-sequence measurements

To process sets of out-of-sequence (OOS) measurements in multi-target tracking, this paper presents an adaptive continuous-discrete trajectory probability hypothesis density filter. The appearances, dynamics and disappearances of targets are considered in the continuous time model, which are discretized by the received in-sequence (IS) or OOS measurements, resulting in the continuous-discrete multi-target model. When receiving OOS measurements, the proposed filter performs retrodiction step through optimal Bayesian processing, and then performs update and marginalisation steps to obtain the best Poisson posterior density approximation on the sets of alive trajectories. Last but not least, A novel and important feature of the filter is to learn the unknown detection profile and clutter rate adaptively at the discrete time steps of measurements.