Weakly Supervised Segmentation of SAR Imagery Using Superpixel and Hierarchically Adversarial CRF
Synthetic aperture radar (SAR) image segmentation aims at generating homogeneous regions from a pixel-based image and is the basis of image interpretation. However, most of the existing segmentation methods usually neglect the appearance and spatial consistency during feature extraction and also req...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2019-03-01
|
Series: | Remote Sensing |
Subjects: | |
Online Access: | http://www.mdpi.com/2072-4292/11/5/512 |
id |
doaj-dc961f65414f4fdfb7caa310484a8534 |
---|---|
record_format |
Article |
spelling |
doaj-dc961f65414f4fdfb7caa310484a85342020-11-24T21:07:30ZengMDPI AGRemote Sensing2072-42922019-03-0111551210.3390/rs11050512rs11050512Weakly Supervised Segmentation of SAR Imagery Using Superpixel and Hierarchically Adversarial CRFFei Ma0Fei Gao1Jinping Sun2Huiyu Zhou3and Amir Hussain4School of Electronic and Information Engineering, Beihang University, Beijing 100191, ChinaSchool of Electronic and Information Engineering, Beihang University, Beijing 100191, ChinaSchool of Electronic and Information Engineering, Beihang University, Beijing 100191, ChinaDepartment of Informatics, University of Leicester, LE1 7RH, UKCognitive Big Data and Cyber-Informatics (CogBID) Laboratory, School of Computing, Edinburgh Napier University, Edinburgh EH10 5DT, Scotland, UKSynthetic aperture radar (SAR) image segmentation aims at generating homogeneous regions from a pixel-based image and is the basis of image interpretation. However, most of the existing segmentation methods usually neglect the appearance and spatial consistency during feature extraction and also require a large number of training data. In addition, pixel-based processing cannot meet the real time requirement. We hereby present a weakly supervised algorithm to perform the task of segmentation for high-resolution SAR images. For effective segmentation, the input image is first over-segmented into a set of primitive superpixels. This algorithm combines hierarchical conditional generative adversarial nets (CGAN) and conditional random fields (CRF). The CGAN-based networks can leverage abundant unlabeled data learning parameters, reducing their reliance on the labeled samples. In order to preserve neighborhood consistency in the feature extraction stage, the hierarchical CGAN is composed of two sub-networks, which are employed to extract the information of the central superpixels and the corresponding background superpixels, respectively. Afterwards, CRF is utilized to perform label optimization using the concatenated features. Quantified experiments on an airborne SAR image dataset prove that the proposed method can effectively learn feature representations and achieve competitive accuracy to the state-of-the-art segmentation approaches. More specifically, our algorithm has a higher Cohen's kappa coefficient and overall accuracy. Its computation time is less than the current mainstream pixel-level semantic segmentation networks.http://www.mdpi.com/2072-4292/11/5/512synthetic aperture radar (SAR)segmentationconditional random fields (CRF)conditional generative adversarial nets (CGAN)neighborhood consistency |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Fei Ma Fei Gao Jinping Sun Huiyu Zhou and Amir Hussain |
spellingShingle |
Fei Ma Fei Gao Jinping Sun Huiyu Zhou and Amir Hussain Weakly Supervised Segmentation of SAR Imagery Using Superpixel and Hierarchically Adversarial CRF Remote Sensing synthetic aperture radar (SAR) segmentation conditional random fields (CRF) conditional generative adversarial nets (CGAN) neighborhood consistency |
author_facet |
Fei Ma Fei Gao Jinping Sun Huiyu Zhou and Amir Hussain |
author_sort |
Fei Ma |
title |
Weakly Supervised Segmentation of SAR Imagery Using Superpixel and Hierarchically Adversarial CRF |
title_short |
Weakly Supervised Segmentation of SAR Imagery Using Superpixel and Hierarchically Adversarial CRF |
title_full |
Weakly Supervised Segmentation of SAR Imagery Using Superpixel and Hierarchically Adversarial CRF |
title_fullStr |
Weakly Supervised Segmentation of SAR Imagery Using Superpixel and Hierarchically Adversarial CRF |
title_full_unstemmed |
Weakly Supervised Segmentation of SAR Imagery Using Superpixel and Hierarchically Adversarial CRF |
title_sort |
weakly supervised segmentation of sar imagery using superpixel and hierarchically adversarial crf |
publisher |
MDPI AG |
series |
Remote Sensing |
issn |
2072-4292 |
publishDate |
2019-03-01 |
description |
Synthetic aperture radar (SAR) image segmentation aims at generating homogeneous regions from a pixel-based image and is the basis of image interpretation. However, most of the existing segmentation methods usually neglect the appearance and spatial consistency during feature extraction and also require a large number of training data. In addition, pixel-based processing cannot meet the real time requirement. We hereby present a weakly supervised algorithm to perform the task of segmentation for high-resolution SAR images. For effective segmentation, the input image is first over-segmented into a set of primitive superpixels. This algorithm combines hierarchical conditional generative adversarial nets (CGAN) and conditional random fields (CRF). The CGAN-based networks can leverage abundant unlabeled data learning parameters, reducing their reliance on the labeled samples. In order to preserve neighborhood consistency in the feature extraction stage, the hierarchical CGAN is composed of two sub-networks, which are employed to extract the information of the central superpixels and the corresponding background superpixels, respectively. Afterwards, CRF is utilized to perform label optimization using the concatenated features. Quantified experiments on an airborne SAR image dataset prove that the proposed method can effectively learn feature representations and achieve competitive accuracy to the state-of-the-art segmentation approaches. More specifically, our algorithm has a higher Cohen's kappa coefficient and overall accuracy. Its computation time is less than the current mainstream pixel-level semantic segmentation networks. |
topic |
synthetic aperture radar (SAR) segmentation conditional random fields (CRF) conditional generative adversarial nets (CGAN) neighborhood consistency |
url |
http://www.mdpi.com/2072-4292/11/5/512 |
work_keys_str_mv |
AT feima weaklysupervisedsegmentationofsarimageryusingsuperpixelandhierarchicallyadversarialcrf AT feigao weaklysupervisedsegmentationofsarimageryusingsuperpixelandhierarchicallyadversarialcrf AT jinpingsun weaklysupervisedsegmentationofsarimageryusingsuperpixelandhierarchicallyadversarialcrf AT huiyuzhou weaklysupervisedsegmentationofsarimageryusingsuperpixelandhierarchicallyadversarialcrf AT andamirhussain weaklysupervisedsegmentationofsarimageryusingsuperpixelandhierarchicallyadversarialcrf |
_version_ |
1716762582487924736 |