Particle Filtering for Target Tracking Using Plot-sequences of Multi-frame Track Before Detect

This paper addresses the tracking problem using the short state sequences in time series (which is called plot-sequences in this paper) obtained from multi-frame track-before-detect (TBD). Some TBD methods, e.g., dynamic programming (DP), maximum likelihood probabilistic data association (ML-PDA), histogram probabilistic multihypothesis tracker (H-PMHT) and Hough transform (HT) based TBD, are batch processing methods whose outputs are plot-sequences which are neither filtered target trajectories nor traditional point detections. In radar systems, traditional tracking algorithms are generally suitable to process detected point plots, which are inappropriate for the plot-sequences. Additionally, many TBD methods are grid based method, which perform the search in the discretized state space. Therefore the estimation accuracy suffers at least half a grid loss. Besides, the missing reports and false alarms in TBD detections will also deprave the tracking results especially when the signal-to-noise ratio (SNR) is low. Thus, a tracker which can combine the plot-sequences into continuous target trajectories that have higher estimation accuracy is needed. In this paper, a nonlinear and non-Gaussian measurement model to describe the plot-sequences of DP based TBD is formulated, then a particle filtering (PF) algorithm for target tracking using the plot-sequences as the input measurements to estimate the target states recursively is developed. Simulation results are presented to demonstrate the effectiveness of the proposed algorithm.