SIViP (2007) 1:149–161
DOI 10.1007/s11760-007-0012-9
ORIGINAL PAPER
Target tracking by fusion of random measures
Mahesh Vemula · Mónica F. Bugallo · Petar M. Djuri´ c
Received: 16 October 2006 / Revised: 23 March 2007 / Accepted: 30 March 2007 / Published online: 25 April 2007
© Springer-Verlag London Limited 2007
Abstract In this paper we propose fusion methods for
tracking a single target in a sensor network. The sensors
use sequential Monte Carlo (SMC) techniques to process the
received measurements and obtain random measures of the
unknown states. We apply standard particle filtering (SPF)
and cost-reference particle filtering (CRPF) methods. For
both types of filtering, the random measures contain parti-
cles drawn from the state space. Associated to the particles,
the SPF has weights representing probability masses, while
the CRPF has user-defined costs measuring the quality of the
particles. Summaries of the random measures are sent to the
fusion center which combines them into a global summary.
Similarly, the fusion center may send a global summary to the
individual sensors that use it for improved tracking. Through
extensive simulations and comparisons with other methods,
we study the performance of the proposed algorithms.
Keywords Multisensor fusion · Target tracking ·
Particle filtering · Cost-reference particle filtering
This work has been supported by the National Science Foundation
under Award CCF-0515246 and the Office of Naval Research under
Award N00014-06-1-0012.
M. Vemula (B )· M. F. Bugallo · P. M. Djuri´ c
Department of Electrical and Computer Engineering,
Stony Brook University, Stony Brook,
NY 11794-2350, USA
e-mail: vema@ece.sunysb.edu
M. F. Bugallo
e-mail: monica@ece.sunysb.edu
P. M. Djuri´ c
e-mail: djuric@ece.sunysb.edu
1 Introduction
Multisensor data fusion refers to the processing and syner-
gistic combination of data from different sensors to provide
improved accuracy and reduced uncertainty about events of
interest [1, 2]. Fusion of data from multiple sensors improves
the robustness and reliability of the system and has a wide
range of application including military, geosciences, robot-
ics, statistical sciences, manufacturing and medicine [2]. In
this paper we study the problem of target tracking by fusion
of information in a sensor network framework as shown in
Fig. 1. There each sensor applies a sequential Monte Car-
lo (SMC) method to obtain a random measure of the target
state. The obtained information is transmitted to the fusion
center (FC), which combines the received data and provides
an estimate of the target state. Clearly, a better performance
could be obtained by transmitting to the FC all the measure-
ments received by the sensors without any processing and
running the SMC method at the FC. However, the transmis-
sion of all these measurements is often not practical, and
therefore we consider local processing at the sensors. More-
over, we assume that the fusion processing occurs periodi-
cally or by request of the FC. A challenge associated to the
proposed scheme comes from the fact that the local SMC
methods produce random measures represented by large sets
of samples and weights/costs. The transmission of the com-
plete measures is therefore prohibitive. We propose solutions
that summarize the random measures and allow for reduced
overall communication load.
The problem of target tracking is usually represented by
using a discrete-time state space (DSS) model. In scenarios
where the noise processes in the DSS are linear and Gauss-
ian, the Kalman filter can be used by the sensors for obtaining
local estimates of the target state [3, 4]. Linear combination
schemes have been proposed in this context to combine local
123