Date of Award

12-1992

Document Type

Thesis

Degree Name

Master of Science in Electrical Engineering

Department

Department of Electrical and Computer Engineering

First Advisor

Dennis Ruck, PhD

Abstract

Fully polarized Xpatch signatures are transformed to two left circularly polarized signals. These two signals are then filtered by a linear FM pulse compression ('chirp') transfer function, corrupted by AWGN, and filtered by a filter matched to the 'chirp' transfer function. The bandwidth of the 'chirp' radar is about 750 MHz. Range profile feature extraction is performed using the TLS Prony Model parameter estimation technique developed at Ohio State University. Using the Prony Model, each scattering center is described by a polarization ellipse, relative energy, frequency response, and range. This representation of the target is vector quantized using a K-means clustering algorithm. Sequences of vector quantized scattering centers as well as sequences of vector quantized range profiles are used to synthesize target specific Hidden Markov Models (HMM's). The identification decision is made by determining which HMM has the highest probability of generating the unknown sequence. The data consist of synthesized Xpatch signatures of two targets which have been difficult to separate with other RTI algorithms. The RTI algorithm developed for this thesis is clearly able to separate these two targets over a 10 by 10 degree (1 degree granularity) aspect angle window off the nose for SNRs as low as 0 dB. The classification rate is 100 % for SNRs of 5 - 20 dB, 95 % for a SNR of 0 dB and it drops rapidly for SNRs lower than 0 dB.

AFIT Designator

AFIT-GE-92-D-15

DTIC Accession Number

ADA259210

Comments

The author's Vita page is omitted.

Download

Share

COinS