Reg-Superpixel Guided Convolutional Neural Network of PolSAR Image Classification Based on Feature Selection and Receptive Field Reconstruction
The convolutional neural network (CNN) has a poor performance in nonuniform and edge regions due to the limitation of fixed receptive field. At the same time, feature stacking of input data can bring burden and overfitting to the network. To solve these problems, this article proposes a reg-superpix...
Main Authors: | , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IEEE
2023-01-01
|
Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
Subjects: | |
Online Access: | https://ieeexplore.ieee.org/document/10110373/ |
_version_ | 1797805978842824704 |
---|---|
author | Ronghua Shang Keyao Zhu Jie Feng Chao Wang Licheng Jiao Songhua Xu |
author_facet | Ronghua Shang Keyao Zhu Jie Feng Chao Wang Licheng Jiao Songhua Xu |
author_sort | Ronghua Shang |
collection | DOAJ |
description | The convolutional neural network (CNN) has a poor performance in nonuniform and edge regions due to the limitation of fixed receptive field. At the same time, feature stacking of input data can bring burden and overfitting to the network. To solve these problems, this article proposes a reg-superpixel guided CNN based on feature selection and receptive field reconstruction. First, a feature selection method is designed, which uses polarimetric SAR statistical distribution features to calculate distance and similarity, and selects features that are easier to identify to avoid the negative impact of low distinguishing features on classification. Second, the reg-superpixel, which means regular superpixel, is used to reconstruct the receptive field and represent the features of the central pixel. The classification result of the central pixel is extended to the whole superpixel during the test. This method can extend the pixel-level CNN network to superpixel-level. Finally, by using edge information of the small-scale superpixel and spatial information of the large-scale superpixel to adjust receptive field of the central pixel, the classification results with uniform smooth region and high edge fitting can be generated. Experimental results with four state-of-the-art methods on four datasets show that feature selection and multiscale reg-superpixel network is effective for polarimetric SAR classification problems. |
first_indexed | 2024-03-13T06:00:05Z |
format | Article |
id | doaj.art-6028bbb3cc8446299f07c7d6e4a70f8b |
institution | Directory Open Access Journal |
issn | 2151-1535 |
language | English |
last_indexed | 2024-03-13T06:00:05Z |
publishDate | 2023-01-01 |
publisher | IEEE |
record_format | Article |
series | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
spelling | doaj.art-6028bbb3cc8446299f07c7d6e4a70f8b2023-06-12T23:00:31ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing2151-15352023-01-01164312432710.1109/JSTARS.2023.326817710110373Reg-Superpixel Guided Convolutional Neural Network of PolSAR Image Classification Based on Feature Selection and Receptive Field ReconstructionRonghua Shang0https://orcid.org/0000-0001-9124-696XKeyao Zhu1https://orcid.org/0000-0002-7219-3695Jie Feng2Chao Wang3https://orcid.org/0000-0001-7578-1970Licheng Jiao4https://orcid.org/0000-0003-3354-9617Songhua Xu5Key Laboratory of Intelligent Perception and Image Understanding of Ministry of Education, School of Artificial Intelligence, Xidian University, Xi'an, ChinaKey Laboratory of Intelligent Perception and Image Understanding of Ministry of Education, School of Artificial Intelligence, Xidian University, Xi'an, ChinaKey Laboratory of Intelligent Perception and Image Understanding of Ministry of Education, School of Artificial Intelligence, Xidian University, Xi'an, ChinaResearch Center for Big Data Intelligence, Zhejiang Laboratory, Hangzhou, ChinaKey Laboratory of Intelligent Perception and Image Understanding of Ministry of Education, School of Artificial Intelligence, Xidian University, Xi'an, ChinaInstitute of Medical Artiffcial Intelligence, The Second Afffliated Hospital of Xi'an Jiaotong University, Xi'an, ChinaThe convolutional neural network (CNN) has a poor performance in nonuniform and edge regions due to the limitation of fixed receptive field. At the same time, feature stacking of input data can bring burden and overfitting to the network. To solve these problems, this article proposes a reg-superpixel guided CNN based on feature selection and receptive field reconstruction. First, a feature selection method is designed, which uses polarimetric SAR statistical distribution features to calculate distance and similarity, and selects features that are easier to identify to avoid the negative impact of low distinguishing features on classification. Second, the reg-superpixel, which means regular superpixel, is used to reconstruct the receptive field and represent the features of the central pixel. The classification result of the central pixel is extended to the whole superpixel during the test. This method can extend the pixel-level CNN network to superpixel-level. Finally, by using edge information of the small-scale superpixel and spatial information of the large-scale superpixel to adjust receptive field of the central pixel, the classification results with uniform smooth region and high edge fitting can be generated. Experimental results with four state-of-the-art methods on four datasets show that feature selection and multiscale reg-superpixel network is effective for polarimetric SAR classification problems.https://ieeexplore.ieee.org/document/10110373/Feature selectionimage classificationpolarimetric decompositionreceptive field remodeling |
spellingShingle | Ronghua Shang Keyao Zhu Jie Feng Chao Wang Licheng Jiao Songhua Xu Reg-Superpixel Guided Convolutional Neural Network of PolSAR Image Classification Based on Feature Selection and Receptive Field Reconstruction IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Feature selection image classification polarimetric decomposition receptive field remodeling |
title | Reg-Superpixel Guided Convolutional Neural Network of PolSAR Image Classification Based on Feature Selection and Receptive Field Reconstruction |
title_full | Reg-Superpixel Guided Convolutional Neural Network of PolSAR Image Classification Based on Feature Selection and Receptive Field Reconstruction |
title_fullStr | Reg-Superpixel Guided Convolutional Neural Network of PolSAR Image Classification Based on Feature Selection and Receptive Field Reconstruction |
title_full_unstemmed | Reg-Superpixel Guided Convolutional Neural Network of PolSAR Image Classification Based on Feature Selection and Receptive Field Reconstruction |
title_short | Reg-Superpixel Guided Convolutional Neural Network of PolSAR Image Classification Based on Feature Selection and Receptive Field Reconstruction |
title_sort | reg superpixel guided convolutional neural network of polsar image classification based on feature selection and receptive field reconstruction |
topic | Feature selection image classification polarimetric decomposition receptive field remodeling |
url | https://ieeexplore.ieee.org/document/10110373/ |
work_keys_str_mv | AT ronghuashang regsuperpixelguidedconvolutionalneuralnetworkofpolsarimageclassificationbasedonfeatureselectionandreceptivefieldreconstruction AT keyaozhu regsuperpixelguidedconvolutionalneuralnetworkofpolsarimageclassificationbasedonfeatureselectionandreceptivefieldreconstruction AT jiefeng regsuperpixelguidedconvolutionalneuralnetworkofpolsarimageclassificationbasedonfeatureselectionandreceptivefieldreconstruction AT chaowang regsuperpixelguidedconvolutionalneuralnetworkofpolsarimageclassificationbasedonfeatureselectionandreceptivefieldreconstruction AT lichengjiao regsuperpixelguidedconvolutionalneuralnetworkofpolsarimageclassificationbasedonfeatureselectionandreceptivefieldreconstruction AT songhuaxu regsuperpixelguidedconvolutionalneuralnetworkofpolsarimageclassificationbasedonfeatureselectionandreceptivefieldreconstruction |