Variational Measurement Update for Extended Object Tracking Using
Gaussian Processes
Abstract
We present an alternative inference framework for the Gaussian
process-based extended object tracking (GPEOT) models. The method
provides an approximate solution to the Bayesian filtering problem in
GPEOT by relying on a new measurement update, which we derive using
variational Bayes techniques. The resulting algorithm effectively
computes approximate posterior densities of the kinematic and the extent
states. We conduct various experiments on simulated and real data and
examine the performance compared with a reference method, which employs
an extended Kalman filter for inference. The proposed algorithm
significantly improves the accuracy of both the kinematic and the extent
estimates and proves robust against model uncertainties.