New subspace methods for ATR

Peng Zhang, Jing Peng, S. Richard F. Sims

Research output: Contribution to journalConference article

Abstract

In ATR applications, each feature is a convolution of an image with a filter. It is important to use most discriminant features to produce compact representations. We propose two novel subspace methods for dimension reduction to address limitations associated with Fukunaga-Koontz Transform (FKT). The first method, Scatter-FKT, assumes that target is more homogeneous, while clutter can be anything other than target and anywhere. Thus, instead of estimating a clutter covariance matrix, Scatter-FKT computes a clutter scatter matrix that measures the spread of clutter from the target mean. We choose dimensions along which the difference in variation between target and clutter is most pronounced. When the target follows a Gaussian distribution, Scatter-FKT can be viewed as a generalization of FKT. The second method, Optimal Bayesian Subspace, is derived from the optimal Bayesian classifier. It selects dimensions such that the minimum Bayes error rate can be achieved. When both target and clutter follow Gaussian distributions, OBS computes optimal subspace representations. We compare our methods against FKT using character image as well as IR data.

Original languageEnglish
Article number39
Pages (from-to)349-358
Number of pages10
JournalProceedings of SPIE - The International Society for Optical Engineering
Volume5807
DOIs
StatePublished - 10 Nov 2005
EventAutomatic Target Recognition XV - Orlando, FL, United States
Duration: 29 Mar 200531 Mar 2005

Fingerprint

Subspace Methods
clutter
Clutter
Mathematical transformations
Transform
Scatter
Target
Gaussian distribution
normal density functions
Subspace
Matrix Measure
Bayesian Classifier
Dimension Reduction
Bayes
classifiers
Covariance matrix
Convolution
Discriminant
convolution integrals
Error Rate

Keywords

  • Bayes method
  • FKT
  • Subspace method

Cite this

@article{bcfdfcb8234a4a01ae46fac83ba079de,
title = "New subspace methods for ATR",
abstract = "In ATR applications, each feature is a convolution of an image with a filter. It is important to use most discriminant features to produce compact representations. We propose two novel subspace methods for dimension reduction to address limitations associated with Fukunaga-Koontz Transform (FKT). The first method, Scatter-FKT, assumes that target is more homogeneous, while clutter can be anything other than target and anywhere. Thus, instead of estimating a clutter covariance matrix, Scatter-FKT computes a clutter scatter matrix that measures the spread of clutter from the target mean. We choose dimensions along which the difference in variation between target and clutter is most pronounced. When the target follows a Gaussian distribution, Scatter-FKT can be viewed as a generalization of FKT. The second method, Optimal Bayesian Subspace, is derived from the optimal Bayesian classifier. It selects dimensions such that the minimum Bayes error rate can be achieved. When both target and clutter follow Gaussian distributions, OBS computes optimal subspace representations. We compare our methods against FKT using character image as well as IR data.",
keywords = "Bayes method, FKT, Subspace method",
author = "Peng Zhang and Jing Peng and Sims, {S. Richard F.}",
year = "2005",
month = "11",
day = "10",
doi = "10.1117/12.604135",
language = "English",
volume = "5807",
pages = "349--358",
journal = "Proceedings of SPIE - The International Society for Optical Engineering",
issn = "0277-786X",
publisher = "SPIE",

}

New subspace methods for ATR. / Zhang, Peng; Peng, Jing; Sims, S. Richard F.

In: Proceedings of SPIE - The International Society for Optical Engineering, Vol. 5807, 39, 10.11.2005, p. 349-358.

Research output: Contribution to journalConference article

TY - JOUR

T1 - New subspace methods for ATR

AU - Zhang, Peng

AU - Peng, Jing

AU - Sims, S. Richard F.

PY - 2005/11/10

Y1 - 2005/11/10

N2 - In ATR applications, each feature is a convolution of an image with a filter. It is important to use most discriminant features to produce compact representations. We propose two novel subspace methods for dimension reduction to address limitations associated with Fukunaga-Koontz Transform (FKT). The first method, Scatter-FKT, assumes that target is more homogeneous, while clutter can be anything other than target and anywhere. Thus, instead of estimating a clutter covariance matrix, Scatter-FKT computes a clutter scatter matrix that measures the spread of clutter from the target mean. We choose dimensions along which the difference in variation between target and clutter is most pronounced. When the target follows a Gaussian distribution, Scatter-FKT can be viewed as a generalization of FKT. The second method, Optimal Bayesian Subspace, is derived from the optimal Bayesian classifier. It selects dimensions such that the minimum Bayes error rate can be achieved. When both target and clutter follow Gaussian distributions, OBS computes optimal subspace representations. We compare our methods against FKT using character image as well as IR data.

AB - In ATR applications, each feature is a convolution of an image with a filter. It is important to use most discriminant features to produce compact representations. We propose two novel subspace methods for dimension reduction to address limitations associated with Fukunaga-Koontz Transform (FKT). The first method, Scatter-FKT, assumes that target is more homogeneous, while clutter can be anything other than target and anywhere. Thus, instead of estimating a clutter covariance matrix, Scatter-FKT computes a clutter scatter matrix that measures the spread of clutter from the target mean. We choose dimensions along which the difference in variation between target and clutter is most pronounced. When the target follows a Gaussian distribution, Scatter-FKT can be viewed as a generalization of FKT. The second method, Optimal Bayesian Subspace, is derived from the optimal Bayesian classifier. It selects dimensions such that the minimum Bayes error rate can be achieved. When both target and clutter follow Gaussian distributions, OBS computes optimal subspace representations. We compare our methods against FKT using character image as well as IR data.

KW - Bayes method

KW - FKT

KW - Subspace method

UR - http://www.scopus.com/inward/record.url?scp=27544483575&partnerID=8YFLogxK

U2 - 10.1117/12.604135

DO - 10.1117/12.604135

M3 - Conference article

AN - SCOPUS:27544483575

VL - 5807

SP - 349

EP - 358

JO - Proceedings of SPIE - The International Society for Optical Engineering

JF - Proceedings of SPIE - The International Society for Optical Engineering

SN - 0277-786X

M1 - 39

ER -