THRESHOLD-BASED RESAMPLING FOR HIGH-SPEED PARTICLE PHD FILTER

In recent years, particle probability hypothesis density (PHD) filtering has become an active research topic for multiple targets tracking in dense clutter scenarios. However, it is highly required to improve the real-time performance of particle PHD filtering because it is a kind of Monte Carlo approach and the computational complexity is very high. One of major di±culties to improve the real- time performance of particle PHD filtering lies in that, resampling, which is usually a sequential process, is crucial to the fully-parallel implementation of particle PHD filter. To overcome this di±culty, this paper presents a novel threshold-based resampling scheme for the particle PHD filter, in which the particle weights are all set below a proper threshold. This specific threshold is determined using a distinguishing feature of the particle PHD filters: The weight sum of all particles in weight update is equal to the total target number in the current iteration. This proposed resampling scheme allows the use of fully-pipelined architecture in the hardware design of particle PHD filter. Theoretical analysis indicates that the particle PHD filter employing the proposed resampling technique can reduce the time complexity by 33% around in a typical multi-target tracking (MTT) scenario compared with that employing the traditional systematic resampling technique, while simulation results show that it can maintain the almost same performance of estimation accuracy.