|
 The
work in this sub-theme is designed to answer specific
questions which will quantify what radar can provide
in the overall process of target identification. Radar
has unique advantages- long range, all-weather operation,
high precision range and Doppler measurement capability-
but the operating wavelengths impose fundamental limits,
especially on angular resolution. Use of other information
(from other sensors, contextual, etc.) can greatly enhance
the recognition capability. A 'multi-functional' capability
of which radar will be a part offers the best prospect
of meeting the operational need, the aim of this research
is to determine what radar can contribute to the whole
picture.
An
important part of the work in this sub-theme is to gain
a better understanding of scattering theories for high-resolution
radar waveforms. The validity of theories, such as Rihaczek
et al, which have challenged conventional assumptions
need to be assessed so that their impact on current
high-resolution radar waveform design and associated
target classification techniques can be determined.
Specific emphasis will be placed on the range resolution
required to achieve a given classifier performance.
It
is recognised that modelling alone will be insufficient
to answer these questions, and that measurements of
real targets are essential. An experimental facility
with 0.6m resolution which can operate at I-band and
F-band is available at BAE SYSTEMS Advanced Technology
Centre and this will be used to support the work.
Existing
identification techniques based on target templates
are not robust. Based on the first stage of the research
described above, the next stage will aim to find alternative
approaches which are less critically dependent on knowledge
of the precise details of the target, as well as knowledge
of its position and orientation during the radar observation.
An important additional aspect of this work is to derive
a robust methodology for providing confidence measures
for the identification process.
This
research theme will also encompass the problem of target
identification in SAR imagery. This problem, particularly
critical to the requirements for rapid large area search,
is aimed at finding targets of interest from a very
large mass of data. Hitherto this has been considered
as an image classification problem, but the complex
radar signature contains much valuable information which
is believed to be exploitable. Certainly, previous work
on coherent change detection has shown the potential
of this approach. This work will concentrate on complex
image analysis as the basis of target detection and
recognition.
|
