Spawning Models for the CPHD Filter

TL;DR

This paper extends the CPHD filter to include spawning models for new targets, providing a more accurate approach in scenarios where targets generate offspring, demonstrated through a Gaussian Mixture implementation and simulation results.

Contribution

It introduces a principled derivation of the CPHD filter with spawning from Finite Set Statistics and presents a Gaussian Mixture implementation.

Findings

Spawning models improve target tracking accuracy.

The proposed filter outperforms birth-only models in simulations.

Multiple spawning scenarios demonstrate the method's effectiveness.

Abstract

In its classical form, the Cardinalized Probability Hypothesis Density (CPHD) filter does not model the appearance of new targets through spawning, yet there are applications for which spawning models more appropriately account for newborn objects when compared to spontaneous birth models. In this paper, we propose a principled derivation of the CPHD filter with spawning from the Finite Set Statistics framework. A Gaussian Mixture implementation of the CPHD filter with spawning is then presented, illustrated with three applicable spawning models on a simulated scenario involving two parent targets spawning a total of five objects. Results show that filter implementations with spawn models provide more accurate results when compared to a birth model implementation.