| In today's technology-driven environment, it is becoming more and
more common for disparate sensors to view the same scene, or at
least a partial overlap of the same scene. Military examples
include missile defense and situation awareness applications,
while non-military related fields include medical imaging and drug
reaction applications. Oftentimes, the view of the scene from
each sensor is passed over communication links either directly to
the other sensors, or to a central 'processor.' Hence, it
is said that the sensors form a network of sorts.
In either case (sensors communicating directly, or to a central
processor), with multiple views of the same scene available, one
important goal is to 'combine', or 'fuse', the
information from the different views to get a (hopefully) more
precise representation of the scene. An important pre-requisite
for fusing multiple views of the same scene is the removal of
systematic sensor registration errors. Systematic sensor
registration errors arise from uncertainty in sensor orientation
(position, pointing angles, internal clock, etc.) and coordinate
transformations, for example. Without properly accounting for
these registration errors, the fused representation of the scene
has the potential to be less precise then any one individual
sensor's view, thus defeating the purpose of a sensor network.
In this paper, we discuss in detail the role of sensor
registration in a sensor network. We present some different
approaches for the registration of sensors, taken from the
literature. In addition, we introduce a new method for sensor
registration based on minimizing a multi-modal energy function. We
compare these different approaches using simulated data. |
| Michael J. Hirsch – biography
Michael earned his Bachelors of Arts degree in mathematics and computer science from West Chester University in 1996 and his Master of Science degree in applied mathematics from the University of Delaware in 1998.
From 1996 to 2001, he worked as a mathematician for AMPAC technologies, a small defense contractor in the Philadelphia area. During this time, Michael worked on computer vision / object recognition problems for the Naval Air Warfare Center – Aircraft Division (NAWC-AD). In July of 2001, Michael joined Raytheon, Inc. He has worked at both the Tucson, Arizona and the St. Petersburg, Florida sites, focusing his research on problems in all areas of data and information fusion.
Since August of 2004, Michael has been a participant in the ‘Raytheon Advanced Study Program’, working full-time on his doctorate in industrial engineering, under Dr. Panos Pardalos at the University of Florida. His thesis work concerns heuristics for continuous non-linear global optimization problems. He expects to defend his dissertation in November of 2006 and graduate in December of 2006.
|