|
| |
Invited Talk
|
|
Combining
SLAM and Visual Surveillance: Problems and Benefits |
|
Dr Ian Reid |
 |
|
Active Vision
Group, University of Oxford, UK |
| |
|
|
|
Simultaneous Localisation and Mapping is the process whereby
a robot or other sensor system builds a map as it moves through a previously
unknown environment, and localize itself relative to the map. In visual SLAM,
the sensor of interest is a camera, and recent years have seen various real-time
visual SLAM systems demonstrated. There might appear to be a duality between the
"inside, looking-out" vision of SLAM, which observes the environment as it moves
through it, and the "outside looking-in" of visual surveillance systems which
typically comprise fixed cameras observing a generally known environment. It is
natural to consider whether this duality can be exploited in positive ways.
Nevertheless there are some important and potentially problematic differences.
Notably, SLAM systems have tended to assume a static environment, while visual
surveillance is more fundamentally |
|
concerned with moving objects or other aspects of a dynamic environment. In this
talk I will discuss two systems in which we seek to exploit SLAM and visual
surveillance to mutual benefit. In the first, we employ a SLAM system within the
context of a multi-PTZ-camera system. We aim, in the first instance, to achieve
a link between the coordinate frames of the SLAM and surveillance systems. More
ambitiously, we aim to achieve cooperative behaviours between the fixed PTZ
cameras and the mobile system; for example to allow an operator equiped with a
SLAM device to "see around corners" via communication with the PTZ system. In
the second application I will describe our work towards SLAM in dynamic
environments, and how this is being used in conjunction with a custom-built high
performance PTZ device to achieve visual contact with targets for marine
situation awareness. |
|
Biography |
|
Ian
Reid is a Reader in Engineering Science and Fellow of Exeter College, at the
University of Oxford where he jointly heads the Active Vision Group. He
obtained BSc from the University of Western Australia in 1987, and came to
Oxford University on a Rhodes Scholarship in 1988 where he completed a
D.Phil. in 1991. His research has touched an many aspects of computer
vision, concentrating on algorithms for visual tracking, control of active
head/eye robotic platforms (for surveillance and navigation), SLAM, visual
geometry, novel view synthesis and human motion capture. He has published
over 100 papers on these and related topics. He has served as an Area Chair
at ECCV and ACCV and currently is on the editorial board of Image and Vision
Computing Journal. |
|
