Introduction
My main research interests are in Computer Vision and Pattern Recognition.
In addition to the projects that are described below, I am open to PhD
project suggestions in any area of Computer Vision and Pattern Recognition,
and I am interested in expanding into areas of vision that involve learning
and biologically inspired vision systems.
Computer Vision for Robot Navigation
Recent work has focused on how to find and segment the ground plane in a
mobile robot's field of view, i.e., which parts of the visible scene can the
robot drive over and which parts must it avoid and treat as obstacles. We
have found a way of doing this with an uncalibrated camera. Furthermore we
have found a way of measuring the height of obstacles above the ground,
again with an uncalibrated camera. This allows us to determine whether
obstacles are small enough to be driven over or high enough to be driven
under. Further work shall focus on how to visually localise a mobile robot
in it's environment while simultaneously building a map.
Computer Vision to Enhance Safety of Civil Aircraft Movements
In this project, we are looking to apply techniques to enhance the safety of
civil aircraft as they move around on runways and taxiways, by highlighting
potential obstacles on a pilot's head-up display. We also wish to consider
the safety implications of using Computer Vision systems in such
application, by working closely with the Department's High Integrity
System's Engineering research group.
Computer Vision for Device Interaction
The aim of this project is to facilitate interactions between two devices,
which we call the "client", such as a mobile phone or personal digital
assistant (PDA), and the "server", such as a personal computer (PC), or
computer controlled public display. An interaction in its simplest form
might be the copying of some data in a file from the server to the client or
vice versa. The way such interactions are controlled in current systems is
usually via custom on-screen menu systems on the server. Other mechanisms
include PDA cradles, where one can press a button on the PDA, when in its
cradle, to initiate a copy of any data changes from the client to the
server. This conventional approach is rather limited, cumbersome and
inflexible. Our approach requires a camera on the client to view the display
on the server, such that either the client or the server (or both) can
compute exactly which part of the server display is being viewed, using
techniques from image processing and computer vision. In addition, either
the client or the server (or both) can compute the six degree-of-freedom (6
DOF) position of the client camera with respect to the server display, so
that we have a "flying mouse". This basic principle will support a very
broad range of interactions (depending on the context in which the system is
operating) and it will make these interactions significantly quicker, easier
and more intuitive for the user to initiate and control.
Face Recognition: 2-D and 3-D Techniques
We are running a face recognition project which has extended many of the
current ideas in 2-D face recognition into 3-D face recognition. 3-D faces
are captured using a special 3-D camera that uses a stereo pair of cameras
and a projector that projects light onto the subject's face. A 3-D mesh can
be generated from the image pair data and a standard 2-D image overlayed as
a texture map. We are looking for the most discriminating feature space
using Linear Discriminant Analysis on a large array of image processing and
feature extraction techniques. Also, we plan to develop novel ways of
representing the 3-D face, which suggest a completely new of implementing
3-D face recognition.
3-D Camera Development: AMADEUS FAR
Associated with the Department's AMADEUS project, we are developing a
standalone face recognition unit, which encapsulates 3-D camera and face
recognition hardware/software in a small, low-power package. The AMADEUS
project aims to develop "architectures, machines and devices for efficient
ubiquitous systems". In this context, we would like to develop a face
recognition unit that can be deployed easily in any home or work space where
user authentication may be required.
AMADEUS Videoware
Again, in the context of small, low-power stand-alone devices, we would like
to build a general purpose architecture on which to implement a wide array
of real-time computer vision algorithms. The kind of applications we are
thinking about include visually driven interaction with devices in the home
(TV, cooker, fridge, etc.), security and surveillance, care of the elderly
and so on. We plan to prototype our architectures using FPGA (field
programmable gate array) technologies.