Digital Signal Processing Reference
In-Depth Information
4.4
Extraction of Dispersive Scattering Features from
Radar Imagery Using Time-Frequency Processing
The joint time-frequency processing of 1D range profiles described in Sec-
tions 4.2 and 4.3 can be extended to deal with 2D radar imagery. ISAR
imaging has long been used by the microwave radar community for object-
diagnostic and target-identification applications. ISAR is a simple and very
robust process for mapping the position and magnitude of the point-scatterers
on a target from multifrequency, multiaspect back-scattered data. However,
for complex targets containing other scattering phenomena such as resonances
and dispersive mechanisms, image artifacts are often encountered in the
resulting ISAR image [27]. One important example is the scattering from
the engine inlet/exhaust duct on aircraft. It is a dominant contributor to
the overall scattering from the target, yet its waveguide-like structure and
the associated frequency-dependent scattering mechanisms make it a non-
point-scattering feature. When processed and displayed by the conventional
ISAR algorithm, the inlet return results in an image feature which is not
well-focused, is not related to the spatial location of the scatterer, and can
often obscure other important point features on the target. Therefore, it
would be useful to automatically remove these artifacts from the ISAR image,
leading to a cleaned ISAR image containing only physically meaningful point-
scatterers. Furthermore, the extracted inlet features can be better displayed in
a more meaningful feature space to identify target resonances and cutoff
phenomena.
Joint time-frequency processing can be applied to ISAR image pro-
cessing to accomplish the above objective [28]. The conceptual idea behind
the joint time-frequency ISAR algorithm is to apply joint time-frequency
transform to the range (or time) axis of the conventional range and cross-
range ISAR image to gain an additional frequency dimension. The result is
a three-dimensional (3D) range, cross-range, and frequency matrix, with
each range and cross-range slice of this matrix representing an ISAR image
at a particular frequency. This concept is illustrated in Figure 4.12. Conse-
quently, by examining how the ISAR image varies with frequency, we can
distinguish the frequency-independent scattering mechanisms from the fre-
quency-dependent ones. In the actual implementation of the joint time-
frequency ISAR, the choice of the joint time-frequency processing engine
is critical to preserve range resolution. This is demonstrated below using the
adaptive Gaussian representation discussed earlier in Sections 2.1.3 and 4.3.4.
The adaptive Gaussian representation has two distinct advantages over
the STFT. First, it is a parametric procedure that results in very high time-
Search WWH ::
Custom Search