Using transfer learning technique for SAR automatic target recognition

Maha Al Mufti; Esra Al Hadhrami; Bilal Taha; Naoufel Werghi

doi:10.1117/12.2538012

Using transfer learning technique for SAR automatic target recognition

Maha Al Mufti
, Esra Al Hadhrami
, Bilal Taha
, Naoufel Werghi

Electrical Engineering

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

3 Scopus citations

Abstract

In this paper, a deep learning approach for Synthetic Aperture Radar (SAR) Automatic Target Recognition (ATR) is proposed. The novelty of the proposed framework stems from the fact that it is based on a transfer learning scheme, where a pre-trained Convolutional Neural Network (CNN) is employed to extract learned features in combination with a classical Support Vector Machine (SVM) for classification. The efficiency of the presented approach is validated on the MSTAR dataset, where ten target classes are used. A classification accuracy of 99.27% is achieved.

Original language	British English
Title of host publication	SPIE Future Sensing Technologies
Editors	Masafumi Kimata, Christopher R. Valenta
Publisher	SPIE
ISBN (Electronic)	9781510631113
DOIs	https://doi.org/10.1117/12.2538012
State	Published - 2019
Event	SPIE Future Sensing Technologies 2019 - Tokyo, Japan Duration: 14 Nov 2019 → …

Publication series

Name	Proceedings of SPIE - The International Society for Optical Engineering
Volume	11197
ISSN (Print)	0277-786X
ISSN (Electronic)	1996-756X

Conference

Conference	SPIE Future Sensing Technologies 2019
Country/Territory	Japan
City	Tokyo
Period	14/11/19 → …

Keywords

Automatic Target Recognition
Deep learning
Remote sensing
Synthetic Aperture Radar
Transfer learning

Access to Document

10.1117/12.2538012

Cite this

@inproceedings{b536375fdbf8410f892404c20107477d,

title = "Using transfer learning technique for SAR automatic target recognition",

abstract = "In this paper, a deep learning approach for Synthetic Aperture Radar (SAR) Automatic Target Recognition (ATR) is proposed. The novelty of the proposed framework stems from the fact that it is based on a transfer learning scheme, where a pre-trained Convolutional Neural Network (CNN) is employed to extract learned features in combination with a classical Support Vector Machine (SVM) for classification. The efficiency of the presented approach is validated on the MSTAR dataset, where ten target classes are used. A classification accuracy of 99.27\% is achieved.",

keywords = "Automatic Target Recognition, Deep learning, Remote sensing, Synthetic Aperture Radar, Transfer learning",

author = "\{Al Mufti\}, Maha and \{Al Hadhrami\}, Esra and Bilal Taha and Naoufel Werghi",

note = "Publisher Copyright: Copyright {\textcopyright} 2019 SPIE.; SPIE Future Sensing Technologies 2019 ; Conference date: 14-11-2019",

year = "2019",

doi = "10.1117/12.2538012",

language = "British English",

series = "Proceedings of SPIE - The International Society for Optical Engineering",

publisher = "SPIE",

editor = "Masafumi Kimata and Valenta, \{Christopher R.\}",

booktitle = "SPIE Future Sensing Technologies",

address = "United States",

}

Al Mufti, M, Al Hadhrami, E, Taha, B & Werghi, N 2019, Using transfer learning technique for SAR automatic target recognition. in M Kimata & CR Valenta (eds), SPIE Future Sensing Technologies., 111970A, Proceedings of SPIE - The International Society for Optical Engineering, vol. 11197, SPIE, SPIE Future Sensing Technologies 2019, Tokyo, Japan, 14/11/19. https://doi.org/10.1117/12.2538012

Using transfer learning technique for SAR automatic target recognition. / Al Mufti, Maha; Al Hadhrami, Esra; Taha, Bilal et al.
SPIE Future Sensing Technologies. ed. / Masafumi Kimata; Christopher R. Valenta. SPIE, 2019. 111970A (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 11197).

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

TY - GEN

T1 - Using transfer learning technique for SAR automatic target recognition

AU - Al Mufti, Maha

AU - Al Hadhrami, Esra

AU - Taha, Bilal

AU - Werghi, Naoufel

PY - 2019

Y1 - 2019

N2 - In this paper, a deep learning approach for Synthetic Aperture Radar (SAR) Automatic Target Recognition (ATR) is proposed. The novelty of the proposed framework stems from the fact that it is based on a transfer learning scheme, where a pre-trained Convolutional Neural Network (CNN) is employed to extract learned features in combination with a classical Support Vector Machine (SVM) for classification. The efficiency of the presented approach is validated on the MSTAR dataset, where ten target classes are used. A classification accuracy of 99.27% is achieved.

AB - In this paper, a deep learning approach for Synthetic Aperture Radar (SAR) Automatic Target Recognition (ATR) is proposed. The novelty of the proposed framework stems from the fact that it is based on a transfer learning scheme, where a pre-trained Convolutional Neural Network (CNN) is employed to extract learned features in combination with a classical Support Vector Machine (SVM) for classification. The efficiency of the presented approach is validated on the MSTAR dataset, where ten target classes are used. A classification accuracy of 99.27% is achieved.

KW - Automatic Target Recognition

KW - Deep learning

KW - Remote sensing

KW - Synthetic Aperture Radar

KW - Transfer learning

UR - https://www.scopus.com/pages/publications/85078169428

U2 - 10.1117/12.2538012

DO - 10.1117/12.2538012

M3 - Conference contribution

AN - SCOPUS:85078169428

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

BT - SPIE Future Sensing Technologies

A2 - Kimata, Masafumi

A2 - Valenta, Christopher R.

PB - SPIE

T2 - SPIE Future Sensing Technologies 2019

Y2 - 14 November 2019

ER -

Using transfer learning technique for SAR automatic target recognition

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this