Detecting Small Moving Targets Based on Probability Hypothesis Density Smoother

摘要

In order to detect and track multiple small targets in low SNR environment, Probability Hypothesis Density (PHD) filter is proposed to solve the problem. When using the PHD filter to track small moving targets in space, the influence of measurement noise may be catastrophic. Measurement noise affects the calculation of particle weights, which results in the estimation error of targets number. This article brings in the concept of smoothing and combines it with PHD filter. When update particle weights, forward recursion and backward smoothing are both used. And at some extent, the influence of measurement noise is weakened. Finally through experiments, compared with standard PHD filter, the algorithm proposed in this article is found to be more accurate in targets number and state estimation.