A new algorithm of SAR image target recognition based on improved deep convolutional neural network

Fei Gao, Teng Huang, Jinping Sun, Jun Wang, Amir Hussain, Erfu Yang

Research output: Contribution to journalArticle

7 Citations (Scopus)
44 Downloads (Pure)

Abstract

Background: To effectively make use of deep learning technology automatic feature extraction ability, and enhance the ability of depth learning method to learn and recognize features, this paper proposed a deep learning algorithm combining Deep Convolutional Neural Network (DCNN) trained with an improved cost function and Support Vector Machine (SVM). Methods: The class separation information, which explicitly facilitates intra-class compactness and interclass separability in the process of learning features, is added to an improved cost function as a regularization term to enhance the feature extraction ability of DCNN. Then the improved DCNN is applied to learn the features of SAR images. Finally, SVM is utilized to map the features into output labels. Results: Experiments are performed on SAR image data in Moving and Stationary Target Acquisition and Recognition (MSTAR) database. The experiment results prove the effectiveness of our method, achieving an average accuracy of 99% on ten types of targets, some variants, and some articulated targets. Conclusion: It proves that our method is effective and CNN enjoys a certain potential to be applied in SAR image target recognition.
Original languageEnglish
Number of pages16
JournalCognitive Computation
Early online date26 Jun 2018
DOIs
Publication statusE-pub ahead of print - 26 Jun 2018

Fingerprint

Aptitude
Learning
Neural networks
Cost functions
Support vector machines
Feature extraction
Costs and Cost Analysis
Learning algorithms
Labels
Experiments
Databases
Technology
Recognition (Psychology)
Deep learning
Support Vector Machine

Keywords

  • synthetic aperture radar (SAR) images
  • automatic target recognition (ATR)
  • deep convolutional neural Network (DCNN)
  • support vector machine (SVM)
  • class separation information

Cite this

@article{748acf6c54a846f09c2a29b687221d9e,
title = "A new algorithm of SAR image target recognition based on improved deep convolutional neural network",
abstract = "Background: To effectively make use of deep learning technology automatic feature extraction ability, and enhance the ability of depth learning method to learn and recognize features, this paper proposed a deep learning algorithm combining Deep Convolutional Neural Network (DCNN) trained with an improved cost function and Support Vector Machine (SVM). Methods: The class separation information, which explicitly facilitates intra-class compactness and interclass separability in the process of learning features, is added to an improved cost function as a regularization term to enhance the feature extraction ability of DCNN. Then the improved DCNN is applied to learn the features of SAR images. Finally, SVM is utilized to map the features into output labels. Results: Experiments are performed on SAR image data in Moving and Stationary Target Acquisition and Recognition (MSTAR) database. The experiment results prove the effectiveness of our method, achieving an average accuracy of 99{\%} on ten types of targets, some variants, and some articulated targets. Conclusion: It proves that our method is effective and CNN enjoys a certain potential to be applied in SAR image target recognition.",
keywords = "synthetic aperture radar (SAR) images, automatic target recognition (ATR) , deep convolutional neural Network (DCNN), support vector machine (SVM), class separation information",
author = "Fei Gao and Teng Huang and Jinping Sun and Jun Wang and Amir Hussain and Erfu Yang",
year = "2018",
month = "6",
day = "26",
doi = "10.1007/s12559-018-9563-z",
language = "English",
journal = "Cognitive Computation",
issn = "1866-9956",

}

A new algorithm of SAR image target recognition based on improved deep convolutional neural network. / Gao, Fei; Huang, Teng; Sun, Jinping; Wang, Jun; Hussain, Amir; Yang, Erfu.

In: Cognitive Computation, 26.06.2018.

Research output: Contribution to journalArticle

TY - JOUR

T1 - A new algorithm of SAR image target recognition based on improved deep convolutional neural network

AU - Gao, Fei

AU - Huang, Teng

AU - Sun, Jinping

AU - Wang, Jun

AU - Hussain, Amir

AU - Yang, Erfu

PY - 2018/6/26

Y1 - 2018/6/26

N2 - Background: To effectively make use of deep learning technology automatic feature extraction ability, and enhance the ability of depth learning method to learn and recognize features, this paper proposed a deep learning algorithm combining Deep Convolutional Neural Network (DCNN) trained with an improved cost function and Support Vector Machine (SVM). Methods: The class separation information, which explicitly facilitates intra-class compactness and interclass separability in the process of learning features, is added to an improved cost function as a regularization term to enhance the feature extraction ability of DCNN. Then the improved DCNN is applied to learn the features of SAR images. Finally, SVM is utilized to map the features into output labels. Results: Experiments are performed on SAR image data in Moving and Stationary Target Acquisition and Recognition (MSTAR) database. The experiment results prove the effectiveness of our method, achieving an average accuracy of 99% on ten types of targets, some variants, and some articulated targets. Conclusion: It proves that our method is effective and CNN enjoys a certain potential to be applied in SAR image target recognition.

AB - Background: To effectively make use of deep learning technology automatic feature extraction ability, and enhance the ability of depth learning method to learn and recognize features, this paper proposed a deep learning algorithm combining Deep Convolutional Neural Network (DCNN) trained with an improved cost function and Support Vector Machine (SVM). Methods: The class separation information, which explicitly facilitates intra-class compactness and interclass separability in the process of learning features, is added to an improved cost function as a regularization term to enhance the feature extraction ability of DCNN. Then the improved DCNN is applied to learn the features of SAR images. Finally, SVM is utilized to map the features into output labels. Results: Experiments are performed on SAR image data in Moving and Stationary Target Acquisition and Recognition (MSTAR) database. The experiment results prove the effectiveness of our method, achieving an average accuracy of 99% on ten types of targets, some variants, and some articulated targets. Conclusion: It proves that our method is effective and CNN enjoys a certain potential to be applied in SAR image target recognition.

KW - synthetic aperture radar (SAR) images

KW - automatic target recognition (ATR)

KW - deep convolutional neural Network (DCNN)

KW - support vector machine (SVM)

KW - class separation information

U2 - 10.1007/s12559-018-9563-z

DO - 10.1007/s12559-018-9563-z

M3 - Article

JO - Cognitive Computation

JF - Cognitive Computation

SN - 1866-9956

ER -