19-01-2013, 11:27 AM
Target Detection Trials with A MilEimeder Wave Radar System
1Target Detection.pdf (Size: 341.75 KB / Downloads: 31)
ABSTRACT
Unobstructed, large RCS targets, similar radar targets
surrounded by moving foliage, and small targets in severe
clutter have been used as test cases for two pre-processing
algorithms and several threshold levels in an experimental
millimeter wave radar system. The rather conventional
"six-out-of-eight" pulse radar selection method with binary
output has been compared to an algorithm that accepts a
target if the pre-defined trigger level is crossed by the
average of the eight consecutive pulses. In this case,
however, the output is an analog value corresponding to the
relative average video amplitude. In terms of plotted video,
this process seems to give a slightly better combination of
false alarm rate and detection probability. Large targets
are easier to detect from foliage clutter with the
conventional method.
INTRODUCTION
Unwanted retums from the ground, vegetation, sea, or
natural objects have been a nuisance to radar designers and
operators as long as there has been radar equipment in use [I].
We take whatever measures available and feasible in order to
reduce the effects of clutter and related returns to target
detectability or to the false alarm rate. In fact, the game is often
a carefully optimized compromise. The advent of
field-operational millimeter wave radars about one decade ago
brought a completely new aspect to this scene. The clutter
environment may depend on the temperature, especially
because water may freeze, come down as snow, and further
cover our radar's near field with a surface of varying radar
cross section. Extensive millimeter wave clutter experiments
and related RCS values, for example, vegetation and
man-made obstacles are documented in [2] and surely form the
basis for any further studies. The complexity of clutter
evaluation at millimeter wave bands is partly due to the tine
spatial resolution that is - at least theoretically - feasible with
these very short wave length systems [3].
TARGETS AND TESTED ALGORITHMS
Three distinctly separate target scenarios were selected for
our experiments. A photographic view of the test range is given
in Figure I . Obviously, the more distant target locations are out
of visual sight. The first target was located at about 400 range
units and had a considerable RCS compared to the surrounding
clutter and obstructing bodies. The obtained raw video plot is
shown in Figure 2, where gray scale color indicates relative
signal amplitude at a specific instant of time and at a defined
range bin. The second target, situated at 280 range units, had
almost similar RCS but was masked by foliage that was
moving due to heavy wind as can he seen in the raw video
sample in Figure 3. Finally, we tested the performance against
a relatively small object whose RCS was of the same order of
magnitude as the surrounding clutter and which was located
much closer to the radar; at about 170 range units. The recorded
raw signal is illustrated in Figure 4.
CONCLUSION
As known by radar operators and designers since decades,
there hardly is a single processing algorithm that gives
adequate detection results in all circumstances and for all
targets. Short-range millimeter wave radars having medium to
high pulse repetition frequencies and utilizing very short pulse
widths (compared to conventional long-range systems) even
without compression do provide interesting alternatives to
signal processing designers, but, at the same time, the adverse
characteristics of surface and volume clutter make the task very
complicated. Our experiments suggest that “brute” detection
algorithms of the form of “ N - ~ ~ t - ~ fmhafy ”n ot give best
results in terms of combined detection probability and false
alarm rate. In most of our test cases, better performance was
obtained by having an “anolog-s@fe”a veraged output used as
the screen intensity input. The main drawback of this method is
the tendency to loose accurate target boundary information.