Group and extended target tracking with the probability hypothesis density filter
Abstract
Multiple target tracking concerns the estimation of an unknown and time-varying
number of objects (targets) as they dynamically evolve over time from a sequence
of measurements obtained from sensors at discrete time intervals. In the Bayesian
ltering framework the estimation problem incorporates natural phenomena such
as false measurements and target birth/death. Though theoretically optimal, the
generally intractable Bayesian lter requires suitable approximations. This thesis
is particularly motivated by a rst-order moment approximation known as the
Probability Hypothesis Density (PHD) lter.
The emphasis in this thesis is on the further development of the PHD lter for
handling more advanced target tracking problems, principally involving multiple
group and extended targets. A group target is regarded as a collection of targets
that share a common motion or characteristic, while an extended target is regarded
as a target that potentially generates multiple measurements.
The main contributions are the derivations of the PHD lter for multiple group
and extended target tracking problems and their subsequent closed-form solutions.
The proposed algorithms are applied in simulated scenarios and their estimate
results demonstrate that accurate tracking performance is attainable for certain
group/extended target tracking problems. The performance is further analysed
with the use of suitable metrics.