An Improved SMC-PHD Filter for Multiple Targets Tracking

Abstract

Two improved contributions have been advanced for the standard Sequential Monte Carlo Probability Hypothesis Density (SMC-PHD) filter. Firstly, a novel method is advanced for the cardinality and state estimation. A weight matrix is firstly calculated by measurements and persistent particles, and the weight sum of each row is then evaluated, the measurements indexed by row will be judged as true if its weight sum is larger than a certain threshold, and the weight sum of persistent particle states will be reported as the true target states. Secondly, an assistant variable which is used to denote the persistent age for every particle is introduced, by the help of this age variable, the overrated problem of targets number in dense clutter environment can be effectively restricted. The results of numerical simulation prove that the improved SMC-PHD filter has higher tracking performance than the standard one.