Global in Local: A Convolutional Transformer for SAR ATR FSL

Wang, Chenwei; Huang, Yulin; Liu, Xiaoyu; Pei, Jifang; Zhang, Yin; Yang, Jianyu

doi:10.1109/lgrs.2022.3183467

Cited by 24 publications

(11 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, several recently proposed self-supervised learning approaches that achieved state-of-the-art results in SAR ATR were selected to be used in comparison experiments. These included three contrastive self-supervised learning models, i.e., the PL method [24], the CDA method [27], and the ConvT method [30]. In addition, four self-supervised learning methods were used, i.e., the TSDF-N method [9], the ICSGF method [10], the SFAS method [42], and the DKTS-N method [43].…”

Section: Comparison With Reference Methodsmentioning

confidence: 99%

“…To increase the classification accuracy of SAR ATR for ships, Xu et al modified the SimSiam framework, which is a classic CSL framework, and developed a new positive pair sampling method that considered polarization information [29]. Wang et al proposed a mixture loss method consisting of contrastive loss and label propagation to investigate the global and local representations in SAR images [30]. Ren et al proposed a Siamese feature embedding network and leveraged the CSL approach to train a low-dimensional feature space for feature extraction in SAR ATR [31].…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Locality Preserving Property Constrained Contrastive Learning for Object Classification in SAR Imagery

et al. 2023

View full text Add to dashboard Cite

Robust unsupervised feature learning is a critical yet tough task for synthetic aperture radar (SAR) automatic target recognition (ATR) with limited labeled data. The developing contrastive self-supervised learning (CSL) method, which learns informative representations by solving an instance discrimination task, provides a novel method for learning discriminative features from unlabeled SAR images. However, the instance-level contrastive loss can magnify the differences between samples belonging to the same class in the latent feature space. Therefore, CSL can dispel these targets from the same class and affect the downstream classification tasks. In order to address this problem, this paper proposes a novel framework called locality preserving property constrained contrastive learning (LPPCL), which not only learns informative representations of data but also preserves the local similarity property in the latent feature space. In LPPCL, the traditional InfoNCE loss of the CSL models is reformulated in a cross-entropy form where the local similarity of the original data is embedded as pseudo labels. Furthermore, the traditional two-branch CSL architecture is extended to a multi-branch structure, improving the robustness of models trained with limited batch sizes and samples. Finally, the self-attentive pooling module is used to replace the global average pooling layer that is commonly used in most of the standard encoders, which provides an adaptive method for retaining information that benefits downstream tasks during the pooling procedure and significantly improves the performance of the model. Validation and ablation experiments using MSTAR datasets found that the proposed framework outperformed the classic CSL method and achieved state-of-the-art (SOTA) results.

show abstract

Section: Comparison With Reference Methodsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Locality Preserving Property Constrained Contrastive Learning for Object Classification in SAR Imagery

et al. 2023

View full text Add to dashboard Cite

show abstract

“…The attention weight is obtained by computing cosine similarity in Euclidean space. Wang et al [33] developed a method combining CNN and transformer, which makes full use of the local perception capability of CNN and the global modeling capability of the transformer. Li et al [34] constructed a multi-aspect SAR sequence dataset from the MSTAR data.…”

Section: Transformer In Target Recognitionmentioning

confidence: 99%

MF-DCMANet: A Multi-Feature Dual-Stage Cross Manifold Attention Network for PolSAR Target Recognition

Zhang

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

The distinctive polarization information of polarimetric SAR (PolSAR) has been widely applied to terrain classification but is rarely used for PolSAR target recognition. The target recognition strategies built upon multi-feature have gained favor among researchers due to their ability to provide diverse classification information. The paper introduces a robust multi-feature cross-fusion approach, i.e., a multi-feature dual-stage cross manifold attention network, namely, MF-DCMANet, which essentially relies on the complementary information between different features to enhance the representation ability of targets. In the first-stage process, a Cross-Feature-Network (CFN) module is proposed to mine the middle-level semantic information of monogenic features and polarization features extracted from the PolSAR target. In the second-stage process, a Cross-Manifold-Attention (CMA) transformer is proposed, which takes the input features represented on the Grassmann manifold to mine the nonlinear relationship between features so that rich and fine-grained features can be captured to compute attention weight. Furthermore, a local window is used instead of the global window in the attention mechanism to improve the local feature representation capabilities and reduce the computation. The proposed MF-DCMANet achieves competitive performance on the GOTCHA dataset, with a recognition accuracy of 99.75%. Furthermore, it maintains a high accuracy rate in the few-shot recognition and open-set recognition scenarios, outperforming the current state-of-the-art method by about 2%.

show abstract

“…In recent decades, deep learning has been applied in signal and image processing fields and demonstrated its superior performance. As for the SAR ATR application, many excellent studies have proposed many deep learning methods with outstanding results [14][15][16][17][18][19][20][21][22][23][24][25]. Chen et al [26] proposed an all-convolutional network replacing all the dense layers with the convolutional layers, which leads to outstanding recognition performance.…”

Section: Introductionmentioning

confidence: 99%

When Deep Learning Meets Multi-Task Learning in SAR ATR: Simultaneous Target Recognition and Segmentation

et al. 2020

Self Cite

View full text Add to dashboard Cite

With the recent advances of deep learning, automatic target recognition (ATR) of synthetic aperture radar (SAR) has achieved superior performance. By not being limited to the target category, the SAR ATR system could benefit from the simultaneous extraction of multifarious target attributes. In this paper, we propose a new multi-task learning approach for SAR ATR, which could obtain the accurate category and precise shape of the targets simultaneously. By introducing deep learning theory into multi-task learning, we first propose a novel multi-task deep learning framework with two main structures: encoder and decoder. The encoder is constructed to extract sufficient image features in different scales for the decoder, while the decoder is a tasks-specific structure which employs these extracted features adaptively and optimally to meet the different feature demands of the recognition and segmentation. Therefore, the proposed framework has the ability to achieve superior recognition and segmentation performance. Based on the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset, experimental results show the superiority of the proposed framework in terms of recognition and segmentation.

show abstract

Global in Local: A Convolutional Transformer for SAR ATR FSL

Cited by 24 publications

References 11 publications

Locality Preserving Property Constrained Contrastive Learning for Object Classification in SAR Imagery

Locality Preserving Property Constrained Contrastive Learning for Object Classification in SAR Imagery

MF-DCMANet: A Multi-Feature Dual-Stage Cross Manifold Attention Network for PolSAR Target Recognition

When Deep Learning Meets Multi-Task Learning in SAR ATR: Simultaneous Target Recognition and Segmentation

Contact Info

Product

Resources

About