The Gaussian sum convolution probability hypothesis density filter

Abstract

A new multi-target tracking algorithm, termed as the Gaussian sum convolution probability hypothesis density (GSCPHD) filter, is proposed. The filter is calculated by a bank of convolution filters with Gaussian approximations to the predicted and posterior densities. It is shown that the ability to deal with complex observation model, non or small observation noise of the GSCPHD over the Gaussian mixture particle PHD (GMPPHD) filter and the lower complexity, more amenable for parallel implementation than the convolution PHD (CPHD) filter. For illustration purposes, the tracking performance of the new filter is presented to compare with the existing GMPPHD filter.