Abstract:
Distributed sensor networks that are tracking the movement of multiple objects
often sense redundant objects due to measurement noises and network errors. This
can cause confusion as the actual number of objects and their actual locations are
difficult to identify. track to track association algorithms can be used to overcome
this problem. Many algorithms have been proposed in literature and they can
be largely classified into two main categories i.e statistical algorithms and clustering based algorithms. One important clustering based algorithm is fuzzy track
to track association algorithm which is the focus of this thesis. A variant of the
fuzzy track to track association algorithm is tested on a set of data generated
through a model that represents a multi sensor multi target scenario environment.
In actual sensors the error in tracks is usually induced in azimuth, elevation and
range values hence an error model based on azimuth, elevation and range system
is proposed in this thesis. The resolutions for the association algorithm are also
based on this realistic error model. In addition, time synchronization of tracks
is also essential before performing track association. A linear predictor that synchronizes the tracks prior to their association is employed in this thesis and the
performance of algorithm is analyzed under time synchronization of tracks. Also a
recent technique based on batch processing is studied which improves the performance of fuzzy track to track association algorithm in certain cases. The technique
in particular is beneficial for the real-time implementation of the algorithm. The
results of our proposed method are compared under various multi sensor multi tarxii
LIST OF FIGURES xiii
get scenarios and it is observed that the proposed method is more than 90 percent
accurately associating the given tracks in scenarios where the noise in tracks is low.
