Training Convolutional Neural Networks for Translational Invariance on SAR ATR

David Malmgren-Hansen; Rasmus  Engholm; Morten  Østergaard Pedersen

Training Convolutional Neural Networks for Translational Invariance on SAR ATR

David Malmgren-Hansen, Rasmus Engholm, Morten Østergaard Pedersen

Terma AS

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

960 Downloads (Pure)

Abstract

In this paper we present a comparison of the robustness of Convolutional Neural Networks (CNN) to other classifiers in the presence of uncertainty of the objects localization in SAR image. We present a framework for simulating simple SAR images, translating the object of interest systematically and testing the classification performance. Our results show that where other classification methods are very sensitive to even small translations, CNN is quite robust to translational variance, making it much more useful in relation to Automatic Target Recognition (ATR) in a real life context.

Original language	English
Title of host publication	Proceedings of EUSAR 2016: 11th European Conference on Synthetic Aperture Radar
Publisher	IEEE
Publication date	2016
Pages	459-462
ISBN (Print)	978-3-8007-4228-8
Publication status	Published - 2016
Event	11th European Conference on Synthetic Aperture Radar - Hamburg, Germany Duration: 6 Jun 2016 → 9 Jun 2016 Conference number: 11 http://conference.vde.com/eusar/2016/Pages/default.aspx

Conference

Conference	11th European Conference on Synthetic Aperture Radar
Number	11
Country/Territory	Germany
City	Hamburg
Period	06/06/2016 → 09/06/2016
Internet address	http://conference.vde.com/eusar/2016/Pages/default.aspx

Access to Document

SARATRBenchmarkAccepted author manuscript, 1.01 MB

OpenUrl availability

Full text

Cite this

@inproceedings{89eae6be7d934b04a7abf4d509820afc,

title = "Training Convolutional Neural Networks for Translational Invariance on SAR ATR",

abstract = "In this paper we present a comparison of the robustness of Convolutional Neural Networks (CNN) to other classifiers in the presence of uncertainty of the objects localization in SAR image. We present a framework for simulating simple SAR images, translating the object of interest systematically and testing the classification performance. Our results show that where other classification methods are very sensitive to even small translations, CNN is quite robust to translational variance, making it much more useful in relation to Automatic Target Recognition (ATR) in a real life context.",

author = "David Malmgren-Hansen and Rasmus Engholm and {{\O}stergaard Pedersen}, Morten",

year = "2016",

language = "English",

isbn = "978-3-8007-4228-8",

pages = "459--462",

booktitle = "Proceedings of EUSAR 2016: 11th European Conference on Synthetic Aperture Radar",

publisher = "IEEE",

address = "United States",

note = "11th European Conference on Synthetic Aperture Radar, EUSAR 2016 ; Conference date: 06-06-2016 Through 09-06-2016",

url = "http://conference.vde.com/eusar/2016/Pages/default.aspx",

}

Training Convolutional Neural Networks for Translational Invariance on SAR ATR. / Malmgren-Hansen, David; Engholm, Rasmus ; Østergaard Pedersen, Morten .
Proceedings of EUSAR 2016: 11th European Conference on Synthetic Aperture Radar. IEEE, 2016. p. 459-462.

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

TY - GEN

T1 - Training Convolutional Neural Networks for Translational Invariance on SAR ATR

AU - Malmgren-Hansen, David

AU - Engholm, Rasmus

AU - Østergaard Pedersen, Morten

N1 - Conference code: 11

PY - 2016

Y1 - 2016

N2 - In this paper we present a comparison of the robustness of Convolutional Neural Networks (CNN) to other classifiers in the presence of uncertainty of the objects localization in SAR image. We present a framework for simulating simple SAR images, translating the object of interest systematically and testing the classification performance. Our results show that where other classification methods are very sensitive to even small translations, CNN is quite robust to translational variance, making it much more useful in relation to Automatic Target Recognition (ATR) in a real life context.

AB - In this paper we present a comparison of the robustness of Convolutional Neural Networks (CNN) to other classifiers in the presence of uncertainty of the objects localization in SAR image. We present a framework for simulating simple SAR images, translating the object of interest systematically and testing the classification performance. Our results show that where other classification methods are very sensitive to even small translations, CNN is quite robust to translational variance, making it much more useful in relation to Automatic Target Recognition (ATR) in a real life context.

M3 - Article in proceedings

SN - 978-3-8007-4228-8

SP - 459

EP - 462

BT - Proceedings of EUSAR 2016: 11th European Conference on Synthetic Aperture Radar

PB - IEEE

T2 - 11th European Conference on Synthetic Aperture Radar

Y2 - 6 June 2016 through 9 June 2016

ER -

Training Convolutional Neural Networks for Translational Invariance on SAR ATR

Abstract

Conference

Access to Document

OpenUrl availability

Fingerprint

Cite this