基于卷积ADMM网络的高效结构化稀疏ISAR成像方法

doi:10.12305/j.issn.1001-506X.2023.01.08

摘要/Abstract

摘要：

结构化稀疏逆合成孔径雷达(inverse synthetic aperture radar, ISAR)成像是空间态势感知与目标识别的重要手段。该问题可通过压缩感知(compressive sensing, CS)方法解决。目前, 许多传统CS方法仍存在运算效率低、参数适应性不强等问题。针对该问题, 本文提出了一种基于卷积交替方向乘子法网络(convolutional alternating direction method of multipliers network, C-ADMMN)的结构化稀疏ISAR成像方法。利用深度展开方法, 结合传统结构化稀疏ISAR成像模型, 构建C-ADMMN网络。通过监督学习, C-ADMMN仅需约10层网络便可达到传统方法上百次迭代的效果, 具有较高的运算效率且对不同目标具有一定适应性。基于仿真与实测数据的实验结果验证了网络的高效性与参数适应性。

关键词: 逆合成孔径雷达, 压缩感知, 深度学习, 深度展开

Abstract:

Structural sparse inverse synthetic aperture radar (ISAR) imaging is an important approach for situation awareness and object recognition. It can be solved via compressive sensing (CS) methods. At present, many conventional CS algorithms still suffer from low computational efficiency and poor parameter adaptability. In this paper, a structural sparse ISAR imaging method based on convolutional alternating direction method of multipliers network (C-ADMMN) is proposed to overcome those problems. The network is established via deep unfolding methods combined with traditional structural sparse ISAR imaging models. The network only needs approximate 10 layers to achieve the effect of hundreds of iterations in traditional methods through supervised learning. The network achieves higher computing efficiency and has a certain sdaptability to different goals, which is proved on the experiment results based on simulated and measured data.

Key words: inverse synthetic aperture radar (ISAR), compressive sensing (CS), deep learning, deep unfolding

中图分类号:

TN957.52

李瑞泽, 张双辉, 刘永祥. 基于卷积ADMM网络的高效结构化稀疏ISAR成像方法[J]. 系统工程与电子技术, 2023, 45(1): 56-70.

Ruize LI, Shuanghui ZHANG, Yongxiang LIU. Computational efficient structural sparse ISAR imaging method based on convolutional ADMM-net[J]. Systems Engineering and Electronics, 2023, 45(1): 56-70.

图/表 29

图1

图2

图3

图4

图5

图6

图7

图8

图9

图10

表1

图11

图12

图13

图14

表2

图15

图16

图17

图18

表3

图19

图20

图21

图22

图23

图24

图25

图26

参考文献 32

1	DONOHOD L.Compressed sensing[J].IEEE Trans.on Information Theory,2006,52(4):1289-1306. doi: 10.1109/TIT.2006.871582
2	ZHANG L, WU S J, DUAN J. Sparse aperture ISAR imaging of maneuvering targets with sparse representation[C]//Proc. of the IEEE Radar Conference, 2015: 1623-1626.
3	ZENGC Z,ZHUW G,JIAX,et al.Sparse aperture ISAR imaging method based on joint constraints of sparsity and low rank[J].IEEE Trans.on Geoscience and Remote Sensing,2021,59(1):168-181. doi: 10.1109/TGRS.2020.2994179
4	TOMEIS,BACCIA,GIUSTIE M,et al.Compressive sensing-based inverse synthetic radar imaging imaging from incomplete data[J].IET Radar, Sonar and Navigation,2016,10(2):386-397. doi: 10.1049/iet-rsn.2015.0290
5	NAZARIM,MEHRPOOYAA,BASTANIM H,et al.High-dimensional sparse recovery using modified generalised SL0 and its application in 3D ISAR imaging[J].IET Radar, Sonar and Navigation,2020,14(8):1267-1278. doi: 10.1049/iet-rsn.2020.0013
6	BOYDS,PARIKHN,CHUE,et al.Distributed optimization and statistical learning via the alternating direction method of multipliers[J].Foundations & Trends in Machine Learning,2010,3(1):1-122.
7	ZHANGS H,LIUY X,LIX.Computationally efficient sparse aperture ISAR autofocusing and imaging based on fast ADMM[J].IEEE Trans.on Geoscience and Remote Sensing,2020,58(12):8751-8765. doi: 10.1109/TGRS.2020.2990445
8	ZHANGS H,LIUY X,LIX.Fast sparse aperture ISAR autofocusing and imaging via ADMM based sparse Bayesian learning[J].IEEE Trans.on Image Processing,2020,29,3213-3226. doi: 10.1109/TIP.2019.2957939
9	邓理康,张双辉,张弛,等.一种基于多维交替方向乘子法的多输入多输出逆合成孔径雷达成像方法[J].雷达学报,2021,10(3):416-431.
	DENGL K,ZHANGS H,ZHANGC.A MIMO-ISAR imaging method based on multi-dimensional ADMM[J].Journal of Radars,2021,10(3):416-431.
10	FANGJ,SHENY N,LIH B,et al.Pattern-coupled sparse Bayesian learning for recovery of block-sparse signals[J].IEEE Trans.on Signal Processng,2015,63(2):360-372. doi: 10.1109/TSP.2014.2375133
11	WANGL,ZHAOL F,RAHARDJAS,et al.Alternative to extended block sparse bayesian learning and its relation to pattern-coupled sparse bayesian learning[J].IEEE Trans.on Signal Processing,2018,66(10):2759-2771. doi: 10.1109/TSP.2018.2816574
12	刘常浩, 卫扬铠, 王岩, 等. 块稀疏贝叶斯学习双基SAR超分辨率成像[C]//第十四届全国信号和智能信息处理与应用学术会议, 2021: 341-347.
	LIU C H, WEI Y K, WANG Y, et al. Bi-SAR super resolution imaging method based on block sparse Bayesian learning[C]//Proc. of the 14th National Conference on Signal and Intelligent Information Processing and Application, 2021: 341-347.
13	FANGJ,ZHANGL Z,LIH B.Two-dimensional pattern-coupled sparse bayesian learning via generalized approximate message passing[J].IEEE Trans.on Image Processing,2016,25(6):2920-2930. doi: 10.1109/TIP.2016.2556582
14	蒲涛,童宁宁,冯为可,等.基于块稀疏矩阵恢复的MIMO雷达扩展目标高分辨成像算法[J].系统工程与电子技术,2021,43(3):647-655.
	PUT,TONGN N,FENGW K,et al.A MIMO radar high resolution imaging algorithm for distributed target based on block sparse matrix recovery[J].Systems Engineering and Electronics,2021,43(3):647-655.
15	ZHANGX H,BAIT,MENGH Y,et al.Compressive sensing-based ISAR imaging via the combination of the sparsity and nonlocal total variation[J].IEEE Geoscience and Remote Sensing Letters,2014,11(5):990-994. doi: 10.1109/LGRS.2013.2284288
16	薛东方,朱晓秀,胡文华,等.基于加权l₁范数优化的双基地ISAR稀疏成像算法[J].系统工程与电子技术,2021,43(4):944-953.
	XUED F,ZHUX X,HUW H,et al.Bi-ISAR sparse imaging algorithm based on reweighted l₁ norm optimization[J].Systems Engineering and Electronics,2021,43(4):944-953.
17	FUS,ZHONGS S,LINL,et al.A novel time-series memory auto-encoder with sequentially updated reconstructions for remaining useful life prediction[J].IEEE Trans.on Neural Networks and Learning Systems, doi: 10.1109/TNNLS.2021.3084249
18	CHENZ,GUOW L,FENGY J,et al.Deep-learned regularization and proximal operator for image compressive sensing[J].IEEE Trans.on Image Processing,2021,30,7112-7126. doi: 10.1109/TIP.2021.3088611
19	XUJ R,YANGY,PUS L,et al.NetRL: task-aware network denoising via deep reinforcement learning[J].IEEE Trans.on Knowledge and Data Engineering, doi: 10.1109/TKDE.2021.3091022
20	SHEIKHOLESLAMIM M,NADIS,NAEINIA A,et al.An efficient deep unsupervised superresolution model for remote sensing images[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2020,13,1937-1945. doi: 10.1109/JSTARS.2020.2984589
21	HERSHEY J R, ROUX J L, WENINGER F. Deep unfolding: model-based inspiration of novel deep architectures[EB/OL]. [2021-07-01]. https://arxiv.org/abs/1409.2574.
22	BORGERDINGM,SCHNITERP,RANGANS.AMP-inspired deep networks for sparse linear inverse problems[J].IEEE Trans.on Signal Processing,2017,65(16):4293-4308. doi: 10.1109/TSP.2017.2708040
23	METZLER C, MOUSAVI A, BARANIUK R. Learned DAMP: principled neural network based compressive image recovery[C]//Proc. of the Advances in Neural Information Processing Systems, 2017: 1772-1783.
24	YANGY,SUNJ,LIH B,et al.ADMM-CSNet: a deep learning approach for image compressive sensing[J].IEEE Trans.on Pattern Analysis and Machine Intelligence,2020,42(3):521-538. doi: 10.1109/TPAMI.2018.2883941
25	LIY Y,CHENGX F,GUIG.Co-robust-ADMM-net: joint ADMM framework and DNN for robust sparse composite regularization[J].IEEE Access,2018,6,47943-47952. doi: 10.1109/ACCESS.2018.2867435
26	WEIX,YANGJ,LVM J,et al.ISAR high-resolution imaging method with joint FISTA and VGGNet[J].IEEE Access,2021,9,86685-86697. doi: 10.1109/ACCESS.2021.3086980
27	WEIS J,LIANGJ D,WANGM,et al.AF-AMPNet: a deep learning approach for sparse aperture ISAR imaging and autofocusing[J].IEEE Trans.on Geoscience and Remote Sensing,2022,60,S206514.
28	HUQ Y,LIUY Z,CAIY L,et al.Joint deep reinforcement learning and unfolding: beam selection and precoding for mmWave multiuser MIMO with lens arrays[J].IEEE Journal on Selected Areas in Communications,2021,39(8):2289-2304. doi: 10.1109/JSAC.2021.3087233
29	汪玲,胡长雨,朱岱寅.基于复数深度神经网络的逆合成孔径雷达成像方法[J].南京航空航天大学学报,2020,52(5):695-700.
	WANGL,HUC Y,ZHUD Y.An ISAR imaging method based on complex-valued deep neural network[J].Journal of Nanjing University of Aeronautics & Astronautics,2020,52(5):695-700.
30	LIR Z,ZHANGS H,ZHANGC,et al.Deep learning approach for sparse aperture isar imaging and autofocusing based on complex-valued ADMM-Net[J].IEEE Sensors Journal,2021,21(3):3437-3451. doi: 10.1109/JSEN.2020.3025053
31	ZHANGS H,LIUY X,LIX,et al.Enhancing ISAR image efficiently via convolutional reweighted l₁ minimization[J].IEEE Trans.on Image Processing,2021,30,4291-4304. doi: 10.1109/TIP.2021.3070442
32	TIBSHIRANIR.Regression shrinkage and selection via the LASSO[J].Journal of the Royal Statistical Society,1996,58(1):267-288.

稀疏度/%	算法	CC	RMSE	PSNR/dB	运算时间/s
50	ADMM^[7]	0.957 0	0.010 2	74.101 7	1.073 2
	CV-ADMMN^[30]	0.979 3	0.006 8	78.467 6	0.343 7
	l₁加权ADMM^[31]	0.975 6	0.007 4	77.464 6	1.846 0
	C-ADMMN	0.979 2	0.006 9	78.626 2	0.069 4
25	ADMM^[7]	0.938 1	0.012 6	72.196 5	4.237 4
	CV-ADMMN^[30]	0.920 9	0.014 3	71.128 7	0.322 9
	l₁加权ADMM^[31]	0.960 9	0.009 5	75.452 0	2.921 9
	C-ADMMN	0.965 5	0.009 0	76.270 0	0.090 0
12.5	ADMM^[7]	0.816 1	0.021 6	66.666 1	3.452 3
	CV-ADMMN^[30]	0.868 3	0.017 5	67.989 5	0.789 1
	l₁加权ADMM^[31]	0.884 9	0.015 7	71.140 7	4.623 7
	C-ADMMN	0.915 1	0.014 27	71.142 2	0.259 8

SNR/dB	算法	CC	RMSE	PSNR/dB	运算时间/s
8	ADMM^[7]	0.935 8	0.012 9	71.809 5	1.226 0
	CV-ADMMN^[30]	0.941 1	0.011 8	72.489 5	0.649 3
	l₁加权ADMM^[31]	0.940 3	0.011 9	73.40 2	3.776 5
	C-ADMMN	0.954 3	0.010 3	74.721 0	0.085 8
0	ADMM^[7]	0.888 0	0.017 0	68.914 7	2.252 8
	CV-ADMMN^[30]	0.898 8	0.016 1	68.762 3	0.675 2
	l₁加权ADMM^[31]	0.937 7	0.012 0	73.768 9	3.466 8
	C-ADMMN	0.947 1	0.011 0	74.561 0	0.069 4
-8	ADMM^[7]	0.708 5	0.022 2	67.179 3	1.538 2
	CV-ADMMN^[30]	0.708 1	0.022 2	67.118 1	0.999 4
	l₁加权ADMM^[31]	0.822 0	0.018 7	71.626 2	5.072 9
	C-ADMMN	0.863 4	0.015 4	72.786 4	0.303 8

稀疏度	算法	CC	RMSE	PSNR/dB	运算时间/s
50%	ADMM^[7]	0.884 0	0.006 2	62.348 1	1.232 1
	CV-ADMMN^[30]	0.915 4	0.005 6	63.136 1	0.676 5
	l₁加权ADMM^[31]	0.909 6	0.005 4	63.921 6	7.915 9
	C-ADMMN	0.948 7	0.004 1	66.186 6	0.172 2
25%	ADMM^[7]	0.861 9	0.006 8	61.561 7	2.918 0
	CV-ADMMN^[30]	0.890 6	0.006 3	61.976 2	0.894 3
	l₁加权ADMM^[31]	0.867 6	0.006 5	61.842 1	7.714 1
	C-ADMMN	0.925 9	0.004 9	64.716 0	0.285 5

[1]	韩啸, 陈世文, 陈蒙, 杨锦程. 基于互易点学习的LPI信号开集识别[J]. 系统工程与电子技术, 2022, 44(9): 2752-2759.
[2]	田鹤, 董纯柱, 殷红成. 基于频域稀疏压缩感知的雷达目标三维散射中心反演方法[J]. 系统工程与电子技术, 2022, 44(9): 2783-2790.
[3]	张立民, 谭凯文, 闫文君, 张聿远. 基于多级跳线残差网络的雷达辐射源识别[J]. 系统工程与电子技术, 2022, 44(7): 2148-2156.
[4]	金国栋, 薛远亮, 谭力宁, 许剑锟. 基于孪生神经网络的目标跟踪算法进展研究[J]. 系统工程与电子技术, 2022, 44(6): 1805-1822.
[5]	徐安林, 张毓, 周峰. 基于Beta过程的高分辨ISAR成像[J]. 系统工程与电子技术, 2022, 44(6): 1873-1879.
[6]	赵晓枫, 徐叶斌, 吴飞, 牛家辉, 蔡伟, 张志利. 基于全局感知机制的地面红外目标检测方法[J]. 系统工程与电子技术, 2022, 44(5): 1461-1467.
[7]	邹虹, 白陈阳, 何鹏, 崔亚平, 王汝言, 吴大鹏. 基于分布式深度学习的边缘服务放置策略[J]. 系统工程与电子技术, 2022, 44(5): 1728-1737.
[8]	刘丰恺, 黄大荣, 郭新荣, 冯存前. 基于吕氏分布的机动目标参数化平动补偿方法[J]. 系统工程与电子技术, 2022, 44(4): 1166-1173.
[9]	陈冬, 句彦伟. 基于语义分割实现的SAR图像舰船目标检测[J]. 系统工程与电子技术, 2022, 44(4): 1195-1201.
[10]	谢艾伦, 刘晓斌, 赵锋, 艾小锋, 肖顺平. 辐射式仿真中PCM信号间歇收发回波重构方法[J]. 系统工程与电子技术, 2022, 44(3): 771-776.
[11]	孙晶明, 虞盛康, 孙俊. 基于深度学习的HRRP识别姿态敏感性分析[J]. 系统工程与电子技术, 2022, 44(3): 802-807.
[12]	姚云翔, 陈莹. 注意力机制下双模态交互融合的目标跟踪网络[J]. 系统工程与电子技术, 2022, 44(2): 410-419.
[13]	高涌荇, 王旭东, 汪玲, 朱岱寅, 郭军, 孟凡旺. 基于RCNN的双极化气象雷达天气信号检测[J]. 系统工程与电子技术, 2022, 44(11): 3380-3387.
[14]	李海, 呼延泽, 毛志杰. 强杂波下基于压缩感知的低空风切变速度解模糊[J]. 系统工程与电子技术, 2022, 44(10): 3029-3036.
[15]	周毅恒, 杨军, 夏赛强, 吕明久. 闪烁现象下旋翼目标微动参数估计方法[J]. 系统工程与电子技术, 2022, 44(1): 54-63.