A novel computationally efficient SMC-PHD Filter using particle-measurement partition

Abstract

The probability hypothesis density (PHD) filter is widely used to solve multi-target tracking (MTT) problems. Although the Sequential Monte Carlo (SMC) implementation provides a tractable solution for PHD filter to handle the highly nonlinear and non-Gaussian MTT scenario, the high computational cost caused by a large number of particles limits the applications that need to be performed in real-time. This paper proposes a computationally efficient SMC-PHD filter using particle-measurement partition and intermediate region strategy. Firstly, the partition strategy provides a way to solve the related PHD calculation in each partition independently. Secondly, based on the rectangular gating technique, the particle intermediate region strategy ensures the estimation accuracy of the proposed method. The simulation results indicate that the partition strategy significantly reduces the computational complexity of the SMC-PHD filter. In addition, the proposed method can maintain comparable accuracy as the standard SMC-PHD filter via the intermediate region strategy.