Implementation of Probability Hypothesis Density Filter with Adaptive Birth Intensity Using Auxiliary Particle Filter

The probability hypothesis density (PHD) filter sequentially computes the first-order multi-target moment for the full multi-target probability density function and dramatically reduces the computational expense of tracking problem. In this paper, we propose an improved implementation of the PHD using the notion of auxiliary particle filter to enhance the effectiveness of the Sequential Monte Carlo (SMC) implementation of the PHD filter. The proposed method differs from traditional SMC implementations because it demonstrates an ability to simultaneously search in an effective way for persistent and newborn targets where the birth intensity is uniform and noninformative. Simulation results indicate that our novel method dramatically improves the accuracy of PHD approximation when compared to traditional SMC implementation methods for the same number of particles.