Collaborative sparse priors for multi-view ATR

Xuelu Li, Vishal Monga

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Recent work has seen a surge of sparse representation based classification (SRC) methods applied to automatic target recognition problems. While traditional SRC approaches used l0 or l1 norm to quantify sparsity, spike and slab priors have established themselves as the gold standard for providing general tunable sparse structures on vectors. In this work, we employ collaborative spike and slab priors that can be applied to matrices to encourage sparsity for the problem of multi-view ATR. That is, target images captured from multiple views are expanded in terms of a training dictionary multiplied with a coefficient matrix. Ideally, for a test image set comprising of multiple views of a target, coefficients corresponding to its identifying class are expected to be active, while others should be zero, i.e. the coefficient matrix is naturally sparse. We develop a new approach to solve the optimization problem that estimates the sparse coefficient matrix jointly with the sparsity inducing parameters in the collaborative prior. ATR problems are investigated on the mid-wave infrared (MWIR) database made available by the US Army Night Vision and Electronic Sensors Directorate, which has a rich collection of views. Experimental results show that the proposed joint prior and coefficient estimation method (JPCEM) can: 1.) enable improved accuracy when multiple views vs. a single one are invoked, and 2.) outperform state of the art alternatives particularly when training imagery is limited.

Original languageEnglish (US)
Title of host publicationAutomatic Target Recognition XXVIII
EditorsAbhijit Mahalanobis, Firooz A. Sadjadi
PublisherSPIE
ISBN (Electronic)9781510618077
DOIs
StatePublished - Jan 1 2018
EventAutomatic Target Recognition XXVIII 2018 - Orlando, United States
Duration: Apr 16 2018Apr 17 2018

Publication series

NameProceedings of SPIE - The International Society for Optical Engineering
Volume10648
ISSN (Print)0277-786X
ISSN (Electronic)1996-756X

Other

OtherAutomatic Target Recognition XXVIII 2018
CountryUnited States
CityOrlando
Period4/16/184/17/18

Fingerprint

Sparsity
Coefficient
coefficients
Sparse Representation
Spike
spikes
Automatic target recognition
slabs
education
Night Vision
Glossaries
night vision
dictionaries
Target
Target Recognition
target recognition
L1-norm
Surge
norms
Gold

All Science Journal Classification (ASJC) codes

  • Electronic, Optical and Magnetic Materials
  • Condensed Matter Physics
  • Computer Science Applications
  • Applied Mathematics
  • Electrical and Electronic Engineering

Cite this

Li, X., & Monga, V. (2018). Collaborative sparse priors for multi-view ATR. In A. Mahalanobis, & F. A. Sadjadi (Eds.), Automatic Target Recognition XXVIII [106480K] (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 10648). SPIE. https://doi.org/10.1117/12.2305387
Li, Xuelu ; Monga, Vishal. / Collaborative sparse priors for multi-view ATR. Automatic Target Recognition XXVIII. editor / Abhijit Mahalanobis ; Firooz A. Sadjadi. SPIE, 2018. (Proceedings of SPIE - The International Society for Optical Engineering).
@inproceedings{10f34ef555c74d1590c803549fc4af65,
title = "Collaborative sparse priors for multi-view ATR",
abstract = "Recent work has seen a surge of sparse representation based classification (SRC) methods applied to automatic target recognition problems. While traditional SRC approaches used l0 or l1 norm to quantify sparsity, spike and slab priors have established themselves as the gold standard for providing general tunable sparse structures on vectors. In this work, we employ collaborative spike and slab priors that can be applied to matrices to encourage sparsity for the problem of multi-view ATR. That is, target images captured from multiple views are expanded in terms of a training dictionary multiplied with a coefficient matrix. Ideally, for a test image set comprising of multiple views of a target, coefficients corresponding to its identifying class are expected to be active, while others should be zero, i.e. the coefficient matrix is naturally sparse. We develop a new approach to solve the optimization problem that estimates the sparse coefficient matrix jointly with the sparsity inducing parameters in the collaborative prior. ATR problems are investigated on the mid-wave infrared (MWIR) database made available by the US Army Night Vision and Electronic Sensors Directorate, which has a rich collection of views. Experimental results show that the proposed joint prior and coefficient estimation method (JPCEM) can: 1.) enable improved accuracy when multiple views vs. a single one are invoked, and 2.) outperform state of the art alternatives particularly when training imagery is limited.",
author = "Xuelu Li and Vishal Monga",
year = "2018",
month = "1",
day = "1",
doi = "10.1117/12.2305387",
language = "English (US)",
series = "Proceedings of SPIE - The International Society for Optical Engineering",
publisher = "SPIE",
editor = "Abhijit Mahalanobis and Sadjadi, {Firooz A.}",
booktitle = "Automatic Target Recognition XXVIII",
address = "United States",

}

Li, X & Monga, V 2018, Collaborative sparse priors for multi-view ATR. in A Mahalanobis & FA Sadjadi (eds), Automatic Target Recognition XXVIII., 106480K, Proceedings of SPIE - The International Society for Optical Engineering, vol. 10648, SPIE, Automatic Target Recognition XXVIII 2018, Orlando, United States, 4/16/18. https://doi.org/10.1117/12.2305387

Collaborative sparse priors for multi-view ATR. / Li, Xuelu; Monga, Vishal.

Automatic Target Recognition XXVIII. ed. / Abhijit Mahalanobis; Firooz A. Sadjadi. SPIE, 2018. 106480K (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 10648).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Collaborative sparse priors for multi-view ATR

AU - Li, Xuelu

AU - Monga, Vishal

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Recent work has seen a surge of sparse representation based classification (SRC) methods applied to automatic target recognition problems. While traditional SRC approaches used l0 or l1 norm to quantify sparsity, spike and slab priors have established themselves as the gold standard for providing general tunable sparse structures on vectors. In this work, we employ collaborative spike and slab priors that can be applied to matrices to encourage sparsity for the problem of multi-view ATR. That is, target images captured from multiple views are expanded in terms of a training dictionary multiplied with a coefficient matrix. Ideally, for a test image set comprising of multiple views of a target, coefficients corresponding to its identifying class are expected to be active, while others should be zero, i.e. the coefficient matrix is naturally sparse. We develop a new approach to solve the optimization problem that estimates the sparse coefficient matrix jointly with the sparsity inducing parameters in the collaborative prior. ATR problems are investigated on the mid-wave infrared (MWIR) database made available by the US Army Night Vision and Electronic Sensors Directorate, which has a rich collection of views. Experimental results show that the proposed joint prior and coefficient estimation method (JPCEM) can: 1.) enable improved accuracy when multiple views vs. a single one are invoked, and 2.) outperform state of the art alternatives particularly when training imagery is limited.

AB - Recent work has seen a surge of sparse representation based classification (SRC) methods applied to automatic target recognition problems. While traditional SRC approaches used l0 or l1 norm to quantify sparsity, spike and slab priors have established themselves as the gold standard for providing general tunable sparse structures on vectors. In this work, we employ collaborative spike and slab priors that can be applied to matrices to encourage sparsity for the problem of multi-view ATR. That is, target images captured from multiple views are expanded in terms of a training dictionary multiplied with a coefficient matrix. Ideally, for a test image set comprising of multiple views of a target, coefficients corresponding to its identifying class are expected to be active, while others should be zero, i.e. the coefficient matrix is naturally sparse. We develop a new approach to solve the optimization problem that estimates the sparse coefficient matrix jointly with the sparsity inducing parameters in the collaborative prior. ATR problems are investigated on the mid-wave infrared (MWIR) database made available by the US Army Night Vision and Electronic Sensors Directorate, which has a rich collection of views. Experimental results show that the proposed joint prior and coefficient estimation method (JPCEM) can: 1.) enable improved accuracy when multiple views vs. a single one are invoked, and 2.) outperform state of the art alternatives particularly when training imagery is limited.

UR - http://www.scopus.com/inward/record.url?scp=85048439332&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85048439332&partnerID=8YFLogxK

U2 - 10.1117/12.2305387

DO - 10.1117/12.2305387

M3 - Conference contribution

AN - SCOPUS:85048439332

T3 - Proceedings of SPIE - The International Society for Optical Engineering

BT - Automatic Target Recognition XXVIII

A2 - Mahalanobis, Abhijit

A2 - Sadjadi, Firooz A.

PB - SPIE

ER -

Li X, Monga V. Collaborative sparse priors for multi-view ATR. In Mahalanobis A, Sadjadi FA, editors, Automatic Target Recognition XXVIII. SPIE. 2018. 106480K. (Proceedings of SPIE - The International Society for Optical Engineering). https://doi.org/10.1117/12.2305387