Dual-branch deep convolution neural network for polarimetric SAR image classification

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

Research output: Contribution to journalArticle

20 Citations (Scopus)

Abstract

Deep convolution neural network (CNN), which has prominent advantages in feature learning, can learn and extract features from data automatically. Existing polarimetric synthetic aperture radar (PolSAR) image classification methods based on CNN only consider the polarization information of the image, instead of incorporating the image’s spatial information. In this paper, a novel method based on dual-branch deep convolution neural network (Dual-CNN) is proposed to realize the classification of PolSAR images. The proposed method is built on two deep CNNs: one is used to extract the polarization features from the 6-channel real matrix (6Ch) which is derived from complex coherency matrix. The other is utilized to extract the spatial features of Pauli RGB image. These extracted features are first combined into a fully connected layer sharing the polarization and spatial property. Then the Softmax classifier is employed to classify these features. The experiments are conducted on the AIRSAR data of Flevoland and the results show that the classification accuracy on 14 types of land cover is up to 98.56%. Such results are promising in comparison with other the state of the art methods.
LanguageEnglish
Article number447
Number of pages18
JournalApplied Sciences
Volume7
Issue number5
Early online date27 Apr 2017
DOIs
Publication statusPublished - 31 May 2017

Fingerprint

image classification
Image classification
Convolution
convolution integrals
Polarization
Synthetic aperture radar
Neural networks
synthetic aperture radar
polarization
Classifiers
matrices
classifiers
learning
Experiments

Keywords

  • deep convolution neural network
  • dual-branch convolution neural network
  • land cover classification
  • polarimetric SAR images

Cite this

Gao, Fei ; Huang, Teng ; Wang, Jun ; Sun, Jinping ; Hussain, Amir ; Yang, Erfu. / Dual-branch deep convolution neural network for polarimetric SAR image classification. In: Applied Sciences. 2017 ; Vol. 7, No. 5.
@article{67bfb384b36c4fdf9645680d285de4f1,
title = "Dual-branch deep convolution neural network for polarimetric SAR image classification",
abstract = "Deep convolution neural network (CNN), which has prominent advantages in feature learning, can learn and extract features from data automatically. Existing polarimetric synthetic aperture radar (PolSAR) image classification methods based on CNN only consider the polarization information of the image, instead of incorporating the image’s spatial information. In this paper, a novel method based on dual-branch deep convolution neural network (Dual-CNN) is proposed to realize the classification of PolSAR images. The proposed method is built on two deep CNNs: one is used to extract the polarization features from the 6-channel real matrix (6Ch) which is derived from complex coherency matrix. The other is utilized to extract the spatial features of Pauli RGB image. These extracted features are first combined into a fully connected layer sharing the polarization and spatial property. Then the Softmax classifier is employed to classify these features. The experiments are conducted on the AIRSAR data of Flevoland and the results show that the classification accuracy on 14 types of land cover is up to 98.56{\%}. Such results are promising in comparison with other the state of the art methods.",
keywords = "deep convolution neural network, dual-branch convolution neural network, land cover classification, polarimetric SAR images",
author = "Fei Gao and Teng Huang and Jun Wang and Jinping Sun and Amir Hussain and Erfu Yang",
year = "2017",
month = "5",
day = "31",
doi = "10.3390/app7050447",
language = "English",
volume = "7",
journal = "Applied Sciences",
issn = "2076-3417",
number = "5",

}

Dual-branch deep convolution neural network for polarimetric SAR image classification. / Gao, Fei; Huang, Teng; Wang, Jun; Sun, Jinping; Hussain, Amir; Yang, Erfu.

In: Applied Sciences, Vol. 7, No. 5, 447, 31.05.2017.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Dual-branch deep convolution neural network for polarimetric SAR image classification

AU - Gao, Fei

AU - Huang, Teng

AU - Wang, Jun

AU - Sun, Jinping

AU - Hussain, Amir

AU - Yang, Erfu

PY - 2017/5/31

Y1 - 2017/5/31

N2 - Deep convolution neural network (CNN), which has prominent advantages in feature learning, can learn and extract features from data automatically. Existing polarimetric synthetic aperture radar (PolSAR) image classification methods based on CNN only consider the polarization information of the image, instead of incorporating the image’s spatial information. In this paper, a novel method based on dual-branch deep convolution neural network (Dual-CNN) is proposed to realize the classification of PolSAR images. The proposed method is built on two deep CNNs: one is used to extract the polarization features from the 6-channel real matrix (6Ch) which is derived from complex coherency matrix. The other is utilized to extract the spatial features of Pauli RGB image. These extracted features are first combined into a fully connected layer sharing the polarization and spatial property. Then the Softmax classifier is employed to classify these features. The experiments are conducted on the AIRSAR data of Flevoland and the results show that the classification accuracy on 14 types of land cover is up to 98.56%. Such results are promising in comparison with other the state of the art methods.

AB - Deep convolution neural network (CNN), which has prominent advantages in feature learning, can learn and extract features from data automatically. Existing polarimetric synthetic aperture radar (PolSAR) image classification methods based on CNN only consider the polarization information of the image, instead of incorporating the image’s spatial information. In this paper, a novel method based on dual-branch deep convolution neural network (Dual-CNN) is proposed to realize the classification of PolSAR images. The proposed method is built on two deep CNNs: one is used to extract the polarization features from the 6-channel real matrix (6Ch) which is derived from complex coherency matrix. The other is utilized to extract the spatial features of Pauli RGB image. These extracted features are first combined into a fully connected layer sharing the polarization and spatial property. Then the Softmax classifier is employed to classify these features. The experiments are conducted on the AIRSAR data of Flevoland and the results show that the classification accuracy on 14 types of land cover is up to 98.56%. Such results are promising in comparison with other the state of the art methods.

KW - deep convolution neural network

KW - dual-branch convolution neural network

KW - land cover classification

KW - polarimetric SAR images

UR - http://www.mdpi.com/journal/applsci

U2 - 10.3390/app7050447

DO - 10.3390/app7050447

M3 - Article

VL - 7

JO - Applied Sciences

T2 - Applied Sciences

JF - Applied Sciences

SN - 2076-3417

IS - 5

M1 - 447

ER -