ARWCGAN:一种高质量的多类别SAR图像生成方法

doi:10.19665/j.issn1001-2400.20250105

摘要/Abstract

摘要：

在合成孔径雷达(SAR)自动目标识别(ATR)领域,高质量的训练数据集通常十分匮乏。现有基于生成对抗网络(GANs)的SAR图像生成方法,常面临训练稳定性差、生成图像质量低的问题。为解决这些问题,提出了一种新的方法,称为注意力残差Wasserstein条件生成对抗网络(ARWCGAN),旨在生成高质量的多类别SAR图像。该方法设计了注意力残差层,以提升模型对SAR图像特征的提取能力,增强生成图像的目标细节和纹理特征。同时,采用了联合梯度惩罚的Wasserstein生成对抗网络(WGAN-GP)损失函数和分类损失函数,以改进训练稳定性并提高生成图像的多样性。在MSTAR数据集进行了生成实验,并从定性视觉检查、定量质量评估和ATR模型贡献三个方面对生成图像效果进行了评估。实验结果表明,ARWCGAN能够生成高质量的图像,显著提升了ATR模型的识别精度。

关键词: 合成孔径雷达, 图像生成, 自动目标识别, 生成对抗网络

Abstract:

In the field of Synthetic Aperture Radar(SAR) Automatic Target Recognition(ATR),the availability of high-quality training datasets is often severely limited.Existing SAR image generation methods based on Generative Adversarial Networks(GANs) suffer from training instability and low-quality outputs.To address these challenges,we propose the Attentional Residual Wasserstein Conditional Generative Adversarial Network(ARWCGAN) for generating high-quality multi-category SAR images.ARWCGAN features attentional residual layers to enhance SAR image feature extraction,thus improving the detail and texture of the generated images.It also utilizes a combined WGAN-GP(Wasserstein Generative Adversarial Network with Gradient Penalty) loss function and classification loss function to improve the training stability and generated image diversity.We conducted generation experiments on the MSTAR dataset and evaluated the generated images from three perspectives:qualitative visual inspection,quantitative quality assessment,and contribution to the ATR model performance.Experimental results demonstrate that ARWCGAN is capable of generating high-quality images,significantly enhancing the recognition accuracy of ATR models.

Key words: Synthetic Aperture Radar(SAR), image generation, automatic target recognition, generative adversarial network

中图分类号:

TP391.41

郑洋, 王榕旭, 郭开泰, 梁继民. ARWCGAN:一种高质量的多类别SAR图像生成方法[J]. 西安电子科技大学学报, 2025, 52(2): 101-112.

ZHENG Yang, WANG Rongxu, GUO Kaitai, LIANG Jimin. ARWCGAN:a method for high-quality multi-category SAR image generation[J]. Journal of Xidian University, 2025, 52(2): 101-112.

图/表 11

图1

图2

图3

表1

图4

图5

表2

表3

表4

图6

表5

参考文献 37

[1]	ZHU X, MONTAZERI S, ALI M, et al. Deep Learning Meets SAR:Concepts,Models,Pitfalls,and Perspectives[J]. IEEE Geoscience and Remote Sensing Magazine, 2021, 9(4):143-172.
[2]	ZHAO Z, JIAO L, ZHAO J, et al. Discriminant Deep Belief Network for High-Resolution SAR Image Classification[J]. Pattern Recognition, 2017, 61:686-701.
[3]	SHAO J, QU C, LI J. A Performance Analysis of Convolutional Neural Network Models in SAR Target Recognition[C]//Proceedings of the 2017 SAR in Big Data Era:Models,Methods and Applications(BIGSARDATA). Piscataway:IEEE, 2017:1-6.
[4]	FENG B, YANG H, ZHANG C, et al. SAR Image Target Recognition Algorithm Based on Convolutional Neural Network[C]//Proceedings of the 2021 IEEE International Conference on Artificial Intelligence and Industrial Design(AIID). Piscataway:IEEE, 2021:364-368
[5]	LI R, WANG X, WANG J, et al. SAR Target Recognition Based on Efficient Fully Convolutional Attention Block CNN[J]. IEEE Geoscience and Remote Sensing Letters, 2022,19:4005905.
[6]	LI S, PAN Z, HU Y. Multi-Aspect Convolutional-Transformer Network for SAR Automatic Target Recognition[J]. Remote Sensing, 2022, 14(16):3924.
[7]	LIU X, HUANG Y, PEI J, et al. Sample Discriminant Analysis for SAR ATR[J]. IEEE Geoscience and Remote Sensing Letters, 2014, 11(12):2120-2124.
[8]	NIU S, QIU X, LEI B, et al. Parameter Extraction Based on Deep Neural Network for SAR Target Simulation[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 58(7):4901-4914.
[9]	AUER S, HINZ S, BAMLER R. Ray-Tracing Simulation Techniques for Understanding High-Resolution SAR Images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2009, 48(3):1445-1456.
[10]	NIU S, LI B, WANG X, et al. Defect Image Sample Generation with GAN for Improving Defect Recognition[J]. IEEE Transactions on Automation Science and Engineering, 2020, 17(3):1611-1622.
[11]	LIU Q, HAO J, GUO Y. EEG Data Augmentation for Emotion Recognition with a Task-Driven GAN[J]. Algorithms, 2023, 16(2):118.
[12]	GUO J, LEI B, DING C, et al. Synthetic Aperture Radar Image Synthesis by Using Generative Adversarial Nets[J]. IEEE Geoscience and Remote Sensing Letters, 2017, 14(7):1111-1115.
[13]	HUANG H, ZHANG F, ZHOU Y, et al. High Resolution SAR Image Synthesis with Hierarchical Generative Adversarial Networks[C]// 2019 IEEE International Geoscience and Remote Sensing Symposium.Piscataway:IEEE, 2019: 2782-2785.
[14]	SCHWEGMANN C P, KLEYNHANS W, SALMON B P, et al. Synthetic Aperture Radar Ship Discrimination,Generation and Latent Variable Extraction Using Information Maximizing Generative Adversarial Networks[C]// 2017 IEEE International Geoscience and Remote Sensing Symposium.Piscataway:IEEE, 2017: 2263-2266.
[15]	DU S, HONG J, WANG Y, et al. A High-Quality Multicategory SAR Images Generation Method with Multiconstraint GAN for ATR[J]. IEEE Geoscience and Remote Sensing Letters, 2021, 19:1-5.
[16]	LIU M, WANG H, CHEN S, et al. A Two-Stage SAR Image Generation Algorithm Based on GAN with Reinforced Constraint Filtering and Compensation Techniques[J]. Remote Sensing, 2024, 16(11):1963.
[17]	ZENG Z, TAN X, ZHANG X, et al. ATGAN:A SAR Target Image Generation Method for Automatic Target Recognition[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024, 17:6290-6307.
[18]	QU Z, FAN G, ZHAO Z, et al. Synthetic Aperture Radar Ground Target Image Generation Based on Improved Wasserstein Generative Adversarial Networks with Gradient Penalty[J]. Journal of Applied Remote Sensing, 2023, 17(3):036501.
[19]	OH J, KIM M. PeaceGAN:A GAN-Based Multi-Task Learning Method for SAR Target Image Generation with a Pose Estimator and an Auxiliary Classifier[J]. Remote Sensing, 2021, 13(19):3939.
[20]	DENG B, GENOVA K, YAZDANI S, et al. Cvxnet:Learnable Convex Decomposition[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. Piscataway:IEEE, 2020:31-44.
[21]	MIRZA M, OSINDERO S. Conditional Generative Adversarial Nets(2014)[J/OL].[2014-11-06]. https://arxiv.org/abs/1411.1784?context=cs.
[22]	GULRAJANI I, AHMED F, ARJOVSKY M, et al. Improved Training of Wasserstein Gans[C]// 31st Conference on Neural Information Processing Systems. Red Hook: NIPS, 2017:5769-5779.
[23]	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is All You Need[C]// 31st Conference on Neural Information Processing Systems. Red Hook: NIPS, 2017:6000-6010.
[24]	ZHANG H, GOODFELLOW I, METAXAS D, et al. Self-Attention Generative Adversarial Networks[C]// International conference on machine learning. New York: PMLR.2019:7354-7363.
[25]	ODENA A, OLAH C, SHLENS J. Conditional Image Synthesis with Auxiliary Classifier Gans[C]// International conference on machine learning. New York: PMLR, 2017:2642-2651.
[26]	MSTAR. MSTAR Dataset(2024)[DS/OL].[2024-01-01]. https://www.sdms.afrl.af.mil/index.php?collection=mstar.
[27]	CAO C, CAO Z, CUI Z. LDGAN:A Synthetic Aperture Radar Image Generation Method for Automatic Target Recognition[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 58(5):3495-3508.
[28]	WANG H, LIU M, CHEN S, et al. Improved SAR Image Generation with Double Top-K Training Method on Auxiliary Classifier GAN[C]// 2023 IEEE International Geoscience and Remote Sensing Symposium.Piscataway:IEEE, 2023: 7046-7049.
[29]	WANG J, LI J, SUN B, et al. SAR Image Synthesis Based on Conditional Generative Adversarial Networks[J]. The Journal of Engineering, 2019(21):8093-8097.
[30]	SALIMANS T, GOODFELLOW I, ZAREMBA W, et al. Improved Techniques for Training Gans[C]// 30th Conference on Neural Information Processing Systems. Red Hook: NIPS, 2016:2234-2242.
[31]	HEUSEL M, RAMSAUER H, UNTERTHINER T, et al. Gans Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium[C]// 31st Conference on Neural Information Processing Systems. Red Hook: NIPS, 2017:6629-6640.
[32]	SZEGEDY C, VANHOUCKE V, IOFFE S, et al. Rethinking the Inception Architecture for Computer Vision[C]// 2016 IEEE Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE, 2016: 2818-2826.
[33]	HE K, ZHANG X, REN S, et al. Deep Residual Learning for Image Recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE, 2016:770-778.
[34]	SIMONYAN K, ZISSERMAN A. Very Deep Convolutional Networks for Large-Scale Image Recognition[C]// International Conference on Learning Representations. Washington DC: ICLR, 2015:1-14.
[35]	DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al. An Image is Worth 16x16 Words:Transformers for Image Recognition at Scale[C]// International Conference on Learning Representations. Washington DC: ICLR, 2021:1-12.
[36]	LIU Z, LIN Y, CAO Y, et al. Swin Transformer:Hierarchical Vision Transformer Using Shifted Windows[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Piscataway:IEEE, 2021:10012-10022.
[37]	HAN K, WANG Y, CHEN H, et al. A Survey on Vision Transformer[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 45(1):87-110.

MSTAR		训练集		ATR测试集
类别	编号	俯仰角	数量	俯仰角	数量
2S1	b01	17°	299	15°	274
BMP2	9566	17°	232	15°	196
BRDM2	E-71	17°	298	15°	274
BTR60	k10yt7532	17°	256	15°	195
BTR70	c71	17°	233	15°	196
D7	92v13015	17°	299	15°	274
T62	A51	17°	299	15°	273
T72	132	17°	232	15°	196
ZIL131	E12	17°	299	15°	274
ZSU23-4	d08	17°	299	15°	274

	SAR-IS↑	SAR-FID↓
ARWCGAN	2.18±0.34	23.32
LDGAN	1.98±0.07	40.13
DC-CGAN	1.85±0.06	56.87

	Inception V3	ResNet101	VGG16	ViT	Swin-T
基线(1∶0)	98.60	98.35	97.40	87.35	96.00
ARWCGAN	98.93	98.75	98.03	88.05	96.35
LDGAN	98.26	97.23	96.79	85.90	92.62
DC-CGAN	97.23	97.47	95.20	78.48	91.46

混合比例	ARWCGAN
基线(1∶0)	87.35
1∶0.5	87.52
1∶1.0	88.05
1∶1.5	89.86
1∶2.0	90.15
1∶2.5	87.96
1∶3.0	87.76

	SAR-IS↑	SAR-FID↓
ARWCGAN	2.18 ± 0.34	23.32
标准线性嵌入	1.12 ± 0.08	95.75
无分类损失	1.91 ± 0.23	46.24