Image Processing Reference
In-Depth Information
-0.05 -0.04 -0.03 -0.02 -0.01
Cross range x (m)
Figure c.3 Imaging buried objects from scaled ISAR experiment (calibration metal sphere and VS-50 plastic
landmine on right). The magnitude is reconstructed from the difference of the ground plate with and without objects
AlgoRIthMS FoR tARget IdentIFICAtIon
Based on the processing steps described above, an estimate of the complex
permittivity distribution in the 3-D domain of interest (either V or a sub-
domain within V ) will be available. Further processing extracts either an
image or signature to identify the target. This step could come before or after
a need to track a suspected target. It is also at this point that multiple images,
taken with two or more radar frequency bands or subapertures, would be
used in conjunction with each other. A goal of this stage is to unambigu-
ously discriminate between decoys, camouflaged structures and nontargets
in order to achieve high recognition rates and low false alarm rates. So, hav-
ing improved the resolution using prior knowledge, the resulting image or
tomographic reconstruction of the image domain can be further processed
to extract specific target information. In the case of SAR imagery, the initial
image resolution is a function of the size of the radiating antenna. In strip
scan SAR as opposed to spotlight mode, the volume V being irradiated is
constantly changing as it moves through the radar footprint. This requires a
modified model for the PDFT to be considered in which a 4-D k -space must
be considered, the fourth dimension representing a weighting applied to the
3-D k -space associated with the sequence of domains V from which scattered
field data are collected.
A Statistical Approach
The PDFT estimates a target such as a discrete image using prior information
to single out one specific estimate from the many possibilities. The approach
we use here with the PDFT is to select a parametrized family of objects
Search WWH ::

Custom Search