Improved Dictionary Formation and Search for Synthetic Aperture Radar Canonical Shape Feature Extraction
Abstract
ATR requires detecting and estimating distinguishing characteristics of a target of interest. Radar data provides range and amplitude information; range distinguishes location relative to the radar whereas amplitude determines strength of re ectivity. Strong re ecting scattering features of targets are detected from a combination of radar returns, or radar PH data. Strong scatterers are modeled as canonical shapes (a plate, dihedral, trihedral, sphere, cylinder, or top-hat). Modeling the scatterers as canonical shapes takes the high dimensional radar PH from each scatterer and parameterizes the scatterer according to its location, size, and orientation. This thesis e ciently estimates the parameters of canonical shapes from radar PH data using dictionary search. Target scattering peaks are detected using 2-D SAR imaging. The parameters are estimated with decreased computation and improved accuracy relative to previous algorithms through reduced SAR image processing, informed parameter subspace bounding, and more e cient dictionary clustering. The e ects of the collection fight path and radar parameters are investigated to permit pre-collection error analysis. The results show that even for a limited collection geometry, the dictionary estimates the canonical shape scatterer parameters well.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 27, 2014
- Accession Number
- ADA598650
Entities
People
- Matthew P. Crosser
Organizations
- Air Force Institute of Technology