Statistical Feature Vector (SFV) For SAR ATR

Historically, Synthetic Aperture Radar (SAR) Automatic Target Recognition (ATR) has faced challenges and performance issues because the tendency is to treat SAR imagery the same way we treat optical imagery, whereas SAR complex images are fundamentally a clutter map that only resembles an image. As...

Full description

Saved in:
Bibliographic Details
Published in2023 IEEE International Radar Conference (RADAR) pp. 1 - 5
Main Authors Picciolo, Michael L., Myrick, Wilbur L., Goldstein, J. Scott
Format Conference Proceeding
LanguageEnglish
Published IEEE 06.11.2023
Subjects
ATR
Automatic Target Recognition
Machine learning algorithms
Optical filters
Optical imaging
radar
Radar imaging
SAR
Scattering
SFV
Statistical Feature Vector
Synthetic Aperture Radar
Target recognition
Wiener Filter
Wiener filters
Online AccessGet full text

Cover

More Information
Summary:Historically, Synthetic Aperture Radar (SAR) Automatic Target Recognition (ATR) has faced challenges and performance issues because the tendency is to treat SAR imagery the same way we treat optical imagery, whereas SAR complex images are fundamentally a clutter map that only resembles an image. As indicated by the name itself, the Statistical Feature Vector (SFV) is a radar phenomenology-based ATR engine that can map many image processing enhancements to the radar domain, while also including radar unique capabilities that can improves SAR ATR. An SFV is based on information theoretic compression of the SAR target formed from a complex Wiener filter that exploits the holographic nature of SAR images. For example, the Multistage Wiener Filter (MWF) may be used that determines the most compressed scattering space (potentially statistical combinations of scattering center elements) that represent the target. The SFV is shown to be a type of sufficient statistic for SAR ATR where methods of employment are the subject of ongoing research.
DOI:10.1109/RADAR54928.2023.10371160