Nan Sun
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Nan Sun has a total of 6 co-authors. Here are the top co-authors by number of shared publications:
- Lin Liu
Department of Computer Science, Xi'an University of Posts and Telecommunications, China
5 shared publications - Wenqiang Hua
Xi’an University of Posts and Telecommunications
5 shared publications - Chen Wang
Department of Computer Science, Xi'an University of Posts and Telecommunications, China
2 shared publications - Xiaomin Jin
School of Computer Science and Technology, Xi’an University of Posts and Telecommunications, Xi’an, China
2 shared publications - Qianjin Hou
Shaanxi Key Laboratory of Network Data Analysis and Intelligent Processing and Xian Key Laboratory of Big Data and Intelligent Computing, School of Computer Science and Technology, Xian University of Posts and Telecommunications, Xian, China
1 shared publication - Zhe Meng
School of Communications and Information Engineering, Xian University of Posts and Telecommunications, Xian, China
1 shared publication
Latest Publications
- 24 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Geoscience and Remote Sensing Letters
- Vol. 21 (1545598X) , 1-5
A Feature Fusion Network for PolSAR Image Classification Based on Physical Features and Deep Features
Deep learning technology has rapidly advanced in the interpretation of polarimetric synthetic aperture radar (PolSAR) images in recent years. However, deep learning methods in PolSAR image interpretation primarily rely on a significant volume of labeled data to make precise predictions, while disregarding the potential physical features of PolSAR. To solve the problem, a deep fusion network is proposed in this letter, which can effectively utilize the complementary characteristics between amplitude and physical features of PolSAR images to enhance the interpretability of the network and improve PolSAR image classification performance. In addition, an improved feature pyramid network and a learnable feature fusion module were proposed to autonomously learn the required fused feature information and avoid the process of feature selection. Finally, the spectral features of PolSAR data are fused to further enhance the discriminability of the features extracted by the proposed network, and improve the classification accuracy of the proposed method. Additionally, to verify the effectiveness of the proposed method, two real PolSAR data sets were used. The experimental results demonstrate that the proposed method achieves higher accuracy, even with a limited number of labeled samples.
- 24 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Geoscience and Remote Sensing Letters
- Vol. 21 (1545598X) , 1-5
A Feature Fusion Network for PolSAR Image Classification Based on Physical Features and Deep Features
Deep learning technology has rapidly advanced in the interpretation of polarimetric synthetic aperture radar (PolSAR) images in recent years. However, deep learning methods in PolSAR image interpretation primarily rely on a significant volume of labeled data to make precise predictions, while disregarding the potential physical features of PolSAR. To solve the problem, a deep fusion network is proposed in this letter, which can effectively utilize the complementary characteristics between amplitude and physical features of PolSAR images to enhance the interpretability of the network and improve PolSAR image classification performance. In addition, an improved feature pyramid network and a learnable feature fusion module were proposed to autonomously learn the required fused feature information and avoid the process of feature selection. Finally, the spectral features of PolSAR data are fused to further enhance the discriminability of the features extracted by the proposed network, and improve the classification accuracy of the proposed method. Additionally, to verify the effectiveness of the proposed method, two real PolSAR data sets were used. The experimental results demonstrate that the proposed method achieves higher accuracy, even with a limited number of labeled samples.
- 2 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Geoscience and Remote Sensing Letters
- Vol. 20 (1545598X) , 1-5
A CA-Based Weighted Clustering Adversarial Network for Unsupervised Domain Adaptation PolSAR Image Classification
With the development of science and technology, although more and more polarimetric synthetic aperture radar (PolSAR) data are collected, marking PolSAR data still requires a lot of costs. Moreover, the datasets between different domains have the class distribution shift problem, which reduces the reusability of labeled samples between cross-domain images. To address this issue, this article proposed an unsupervised domain adaptation (UDA) network based on coordinate attention (CA) and weighted clustering. Firstly, an adversarial UDA network with a bi-classifier is introduced to eliminate the problem of class distribution shift and achieve alignment of data distribution between different domains. Secondly, the CA mechanism is introduced to select important features to enhance the utilization of spatial information among pixels. Finally, to improve the utilization of semantic and classification information of the target domain, and to align same class samples, a weighted clustering algorithm is introduced. Experimental results show that compared with the existing UDA method, the proposed method can achieve the better PolSAR image classification.
- 2 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Geoscience and Remote Sensing Letters
- Vol. 20 (1545598X) , 1-5
A CA-Based Weighted Clustering Adversarial Network for Unsupervised Domain Adaptation PolSAR Image Classification
With the development of science and technology, although more and more polarimetric synthetic aperture radar (PolSAR) data are collected, marking PolSAR data still requires a lot of costs. Moreover, the datasets between different domains have the class distribution shift problem, which reduces the reusability of labeled samples between cross-domain images. To address this issue, this article proposed an unsupervised domain adaptation (UDA) network based on coordinate attention (CA) and weighted clustering. Firstly, an adversarial UDA network with a bi-classifier is introduced to eliminate the problem of class distribution shift and achieve alignment of data distribution between different domains. Secondly, the CA mechanism is introduced to select important features to enhance the utilization of spatial information among pixels. Finally, to improve the utilization of semantic and classification information of the target domain, and to align same class samples, a weighted clustering algorithm is introduced. Experimental results show that compared with the existing UDA method, the proposed method can achieve the better PolSAR image classification.
- Open Access
- 22 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 194-199
Unsupervised Cross-Domain Network Based on Bi-classifier for PolSAR Image Classification
Polarimetric Synthetic Aperture Radar (PolSAR) image classification is an important research content in the field of remote sensing. However, the lack of effective labeled data and the serious class distribution shift problem between different datasets hinder the application of deep learning networks in PolSAR image classification. Therefore, this paper proposes a Bi-classifier adversarial network based on covariance difference matrix and Convolutional Block Attention Module (CBAM) to solve the cross-domain problem in PolSAR images. Firstly, since PolSAR image processing generally uses the covariance matrix or coherence matrix of a single target pixel to construct the data set, which cannot effectively use the local information of PolSAR data,Therefore, the PolSAR coherence difference matrix P is proposed as the description of the PolSAR image difference information. Secondly, due to the different importance of features between neighboring pixels, the CBAM mechanism is introduced in order to exploit the spatial information of PolSAR data. Finally, in order to solve the cross-domain problem of PolSAR images more effectively, a bi-classifier adversarial network based on covariance difference matrix and CBAM is proposed. In this paper, we conduct experiments on public datasets, using images of San FranciscoⅠ acquired in 2008 as the source domain, and images of San Francisco Ⅱ acquired in 1992 as the target domain data. The experimental results show that the proposed method can effectively solve the cross-domain problem of PolSAR images.
- 22 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 194-199
Open AccessUnsupervised Cross-Domain Network Based on Bi-classifier for PolSAR Image Classification
Polarimetric Synthetic Aperture Radar (PolSAR) image classification is an important research content in the field of remote sensing. However, the lack of effective labeled data and the serious class distribution shift problem between different datasets hinder the application of deep learning networks in PolSAR image classification. Therefore, this paper proposes a Bi-classifier adversarial network based on covariance difference matrix and Convolutional Block Attention Module (CBAM) to solve the cross-domain problem in PolSAR images. Firstly, since PolSAR image processing generally uses the covariance matrix or coherence matrix of a single target pixel to construct the data set, which cannot effectively use the local information of PolSAR data,Therefore, the PolSAR coherence difference matrix P is proposed as the description of the PolSAR image difference information. Secondly, due to the different importance of features between neighboring pixels, the CBAM mechanism is introduced in order to exploit the spatial information of PolSAR data. Finally, in order to solve the cross-domain problem of PolSAR images more effectively, a bi-classifier adversarial network based on covariance difference matrix and CBAM is proposed. In this paper, we conduct experiments on public datasets, using images of San FranciscoⅠ acquired in 2008 as the source domain, and images of San Francisco Ⅱ acquired in 1992 as the target domain data. The experimental results show that the proposed method can effectively solve the cross-domain problem of PolSAR images.
- Open Access
- 22 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 200-204
Polarimetric SAR Images Classification Based on Convolutional Neural Networks
Polarimetric synthetic aperture radar (PolSAR) image classification is a dynamic and important topic in remote sensing image interpretation. However, most deep learning models require enough labeled samples to achieve satisfactory performance, and it is difficult to obtain a large number of labeled samples in practice. To alleviate the problem of insufficient label information, we introduce a contrastive learning method for PolSAR image classification. First, we select a certain number of unlabeled samples in the PolSAR data, and then apply an augmentation strategy to generate positive and negative pairs. Then, the pre-trained model learns representations that are beneficial for downstream tasks from these positive and negative samples. Finally, we extract pretrained weights from this network and fine-tune the linear model for classification using a small number of labeled samples. In addition, we introduce a feature extraction method that enhances the representation of each image by incorporating spatial pyramid pooling. Experiments show that our network works by obtaining improved results with a small amount of labeled data.
- 22 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 200-204
Open AccessPolarimetric SAR Images Classification Based on Convolutional Neural Networks
Polarimetric synthetic aperture radar (PolSAR) image classification is a dynamic and important topic in remote sensing image interpretation. However, most deep learning models require enough labeled samples to achieve satisfactory performance, and it is difficult to obtain a large number of labeled samples in practice. To alleviate the problem of insufficient label information, we introduce a contrastive learning method for PolSAR image classification. First, we select a certain number of unlabeled samples in the PolSAR data, and then apply an augmentation strategy to generate positive and negative pairs. Then, the pre-trained model learns representations that are beneficial for downstream tasks from these positive and negative samples. Finally, we extract pretrained weights from this network and fine-tune the linear model for classification using a small number of labeled samples. In addition, we introduce a feature extraction method that enhances the representation of each image by incorporating spatial pyramid pooling. Experiments show that our network works by obtaining improved results with a small amount of labeled data.
- Open Access
- 22 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 212-217
Semi-supervised PolSAR image classification method based on contrastive learning
Polarimetric synthetic aperture radar (PolSAR) image classification has important application value and a wide range of application scenarios in many fields. Supervised classification methods, which need to use a large amount of labeled data to improve classification accuracy, have achieved satisfactory classification performance. However, it is difficult and expensive to obtain sufficient labeled data for PolSAR images. The problem to be solved is how to obtain high classification accuracy with a limited number of labeled samples. Semi-supervised PolSAR image classification method based on contrastive learning is proposed. Firstly, a supervised PolSAR image classification method based on meta-learning is introduced. Prototypical network based on meta-learning ideas is used as the basic classification model, and a prototypical representation is created for each category. Classification is performed by computing the similarity between the prototypical representation and samples to be classified. Secondly, a method of training feature extractor based on contrastive learning is designed. By maintaining the consistency of features between the same data and distinguishing the differences between different data after data augmentation, the method enables the feature extractor to obtain accurate feature representation. By co-training with meta-learning, so as to improve the classification performance. Finally, experiments on real PolSAR data prove the effectiveness of the proposed method.
- 22 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 212-217
Open AccessSemi-supervised PolSAR image classification method based on contrastive learning
Polarimetric synthetic aperture radar (PolSAR) image classification has important application value and a wide range of application scenarios in many fields. Supervised classification methods, which need to use a large amount of labeled data to improve classification accuracy, have achieved satisfactory classification performance. However, it is difficult and expensive to obtain sufficient labeled data for PolSAR images. The problem to be solved is how to obtain high classification accuracy with a limited number of labeled samples. Semi-supervised PolSAR image classification method based on contrastive learning is proposed. Firstly, a supervised PolSAR image classification method based on meta-learning is introduced. Prototypical network based on meta-learning ideas is used as the basic classification model, and a prototypical representation is created for each category. Classification is performed by computing the similarity between the prototypical representation and samples to be classified. Secondly, a method of training feature extractor based on contrastive learning is designed. By maintaining the consistency of features between the same data and distinguishing the differences between different data after data augmentation, the method enables the feature extractor to obtain accurate feature representation. By co-training with meta-learning, so as to improve the classification performance. Finally, experiments on real PolSAR data prove the effectiveness of the proposed method.