基于小波散射激励与SimAM的轻量化抗噪卷积神经网络的目标识别

doi:10.12305/j.issn.1001-506X.2026.07.06

系统工程与电子技术 ›› 2026, Vol. 48 ›› Issue (7): 2184-2194.doi: 10.12305/j.issn.1001-506X.2026.07.06

• 传感器与信号处理 • 上一篇

基于小波散射激励与SimAM的轻量化抗噪卷积神经网络的目标识别

刘畅¹(), 王凌宇¹(), 夏浪¹, 李岳峰², 林欣³, 刘艳阳³, 黄鹏辉¹

1. 上海交通大学集成电路学院（信息与电子工程学院），上海 200240
2. 海军航空大学信息融合研究所，山东烟台 264001
3. 上海卫星工程研究所，上海 201109

收稿日期:2025-04-21 修回日期:2025-08-02 接受日期:2025-08-12 出版日期:2025-11-25 发布日期:2025-11-25
通讯作者: 王凌宇 E-mail:liuchang2024@sjtu.edu.cn;wly123@sjtu.edu.cn
基金资助:
国家自然科学基金（62171272）；中国航天科技集团公司第八研究院产学研合作基金（USCAST2022-21，USCAST2023-29）；东方人才计划青年项目（CYQN2023076）资助课题

Target recognition of lightweight noise-resistant CNN based on wavelet scattering excitation and SimAM

Chang LIU¹(), Lingyu WANG¹(), Lang XIA¹, Yuefeng LI², Xin LIN³, Yanyang LIU³, Penghui HUANG¹

1. School of Integrated Circuits （School of Information Science and Electronic Engineering），Shanghai Jiao Tong University，Shanghai 200240，China
2. Institute of Information Fusion，Naval Aviation University，Yantai 264001，China
3. Shanghai Satellite Engineering Research Institute，Shanghai 201109，China

Received:2025-04-21 Revised:2025-08-02 Accepted:2025-08-12 Online:2025-11-25 Published:2025-11-25
Contact: Lingyu WANG E-mail:liuchang2024@sjtu.edu.cn;wly123@sjtu.edu.cn

摘要/Abstract

摘要：

针对合成孔径雷达图像识别网络消耗部署资源大和散斑噪声强的问题，提出一种基于小波散射激励与简单注意力模块（wavelet scattering excitation and simple attention module，WSS）的轻量化抗噪卷积神经网络（convolutional neural network, CNN）雷达目标识别方法。首先，使用WSS模块让模型对噪声图像进行自主学习，使模型逐渐关注飞机散射中心，从而减少噪声干扰。其次，利用简单注意力模块（simple attention module，SimAM）进行进一步抗噪，并动态计算每个位置的特征图。最后，利用全连接层进行分类输出，在实现参数量大幅度降低的同时具有较高的识别准确性能。不同噪声条件下的实验结果表明，与现有网络模型相比，WSS-CNN在SAR图像识别任务中具有更高的准确率，在复杂噪声环境下展现出优越性能。

关键词: 合成孔径雷达, 卷积神经网络, 散斑噪声, 小波散射激励, 简单注意力模块

Abstract:

Aiming at the problems of large deployment resources consumption and strong speclcle noise of synthetic aperture radar image recognition network, a lightweight noise-resistant convolutional neural network （CNN） based on wavelet scattering excitation and simple attention module （WSS） for radar target recognition method is proposed. Firstly, the noise interference is reduced by using the WSS module to allow the model to learn autonomously from the noisy image so that the model gradually focuses on the scattering center of the aircraft. Secondly, the simple attention module （SimAM） is utilized for further noise immunity and the feature maps are dynamically computed for each position. Finally, the full connectivity layer is utilized for the classification output, which has high recognition accuracy performance while achieving a significant reduction in the number of parameters. The experimental results under different noise conditions show that compared with the existing network models, WSS-CNN achieves higher accuracy in the SAR image recognition task, and it has superior performance in complex noise environments.

Key words: synthetic aperture radar （SAR）, convolutional neural network （CNN）, speckle noise, wavelet scattering excitation, simple attention module （SimAM）

中图分类号:

TN 95

刘畅, 王凌宇, 夏浪, 李岳峰, 林欣, 刘艳阳, 黄鹏辉. 基于小波散射激励与SimAM的轻量化抗噪卷积神经网络的目标识别[J]. 系统工程与电子技术, 2026, 48(7): 2184-2194.

Chang LIU, Lingyu WANG, Lang XIA, Yuefeng LI, Xin LIN, Yanyang LIU, Penghui HUANG. Target recognition of lightweight noise-resistant CNN based on wavelet scattering excitation and SimAM[J]. Systems Engineering and Electronics, 2026, 48(7): 2184-2194.

图/表 17

图1

图2

图3

图4

表1

表2

表3

图5

图6

表4

表5

表6

表7

图7

图8

表8

图9

参考文献 39

1	LEE S, JOO J, CHOI J, et al. W-band multichannel FMCW radar sensor with switching-TX antennas[J]. IEEE Sensors Journal, 2016, 16 (14): 5572- 5582. doi: 10.1109/JSEN.2016.2567450
2	CHEN J L, XIONG R Q, YU H W, et al. Nonparametric full-aperture autofocus imaging for microwave photonic SARs[J]. IEEE Trans. on Geoscience and Remote Sensing, 2024, 62, 5214815.
3	OHMORI T, KIDERA S. Doppler velocity enhanced range migration algorithm for high resolution and noise-robust three-dimensional radar imagings[J]. IEEE Sensors Journal, 2021, 21 (18): 20616- 20628. doi: 10.1109/JSEN.2021.3098322
4	CHEN Y J, WANG S Y, LUO Y, et al. Measurement matrix optimization based on target prior information for radar imagings[J]. IEEE Sensors Journal, 2023, 23 (9): 9808- 9819. doi: 10.1109/JSEN.2023.3263591
5	CHEN L, JIANG X, LI Z, et al. Feature-enhanced speckle reduction via low-rank and space-angle continuity for circular SAR target recognitions[J]. IEEE Trans. on Geoscience and Remote Sensing, 2020, 58 (11): 7734- 7752. doi: 10.1109/TGRS.2020.2983420
6	NOVAK L M, OWIRKA G J, BROWER W S. Performance of 10- and 20-target MSE classifierss[J]. IEEE Trans. on Aerospace and Electronic Systems, 2000, 36 (4): 1279- 1289. doi: 10.1109/7.892675
7	EL-DARYMLI K, MCGUIRE P, GILL E W, et al. Holism-based features for target classification in focused and complex-valued synthetic aperture radar imagerys[J]. IEEE Trans. on Aerospace and Electronic Systems, 2016, 52 (2): 786- 808. doi: 10.1109/TAES.2015.140757
8	SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognitions[EB/OL]. [2025-04-21]. https://arxiv.org/abs/1409.1556.
9	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognitions[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770−778.
10	HOWARD A G. MobileNets: efficient convolutional neural networks for mobile vision applicationss[EB/OL]. [2025-04-21]. https://arxiv.org/abs/1704.04861.
11	DENG J, DONG W, SOCHER R, et al. ImageNet: a large-scale hierarchical image databases[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2009: 248−255.
12	MA N, ZHANG X, ZHENG H T, et al. ShuffleNet V2: practical guidelines for efficient CNN architecture designs[C]//Proc. of the European Conference on Computer Vision, 2018: 116−131.
13	CHEN S Z, WANG H P, XU F, et al. Target classification using the deep convolutional networks for SAR imagess[J]. IEEE Trans. on Geoscience and Remote Sensing, 2016, 54 (8): 4806- 4817. doi: 10.1109/TGRS.2016.2551720
14	LECUN Y, BOTTOU L, BENGIO Y, et al. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998, 86(11): 2278−2324.
15	王彩云, 吴钇达, 王佳宁, 等. 基于改进的CNN和数据增强的SAR目标识别[J]. 系统工程与电子技术, 2022, 44 (8): 2483- 2487. doi: 10.12305/j.issn.1001-506X.2022.08.12
	WANG C Y, WU Y D, WANG J N, et al. SAR image target recognition based on combinatorial optimization convolutional neural network[J]. Systems Engineering and Electronics, 2022, 44 (8): 2483- 2487. doi: 10.12305/j.issn.1001-506X.2022.08.12
16	蒋明煜, 张顺生, 肖思瑶. 基于轻量级交叉注意力卷积网络的SAR目标识别[J]. 系统工程与电子技术, 2025, 47 (9): 2853- 2861.
	JIANG M Y, ZHANG S S, XIAO S Y. SAR target recognition based on lightweight cross-attention convolutional network[J]. Systems Engineering and Electronics, 2025, 47 (9): 2853- 2861.
17	WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention modules[C]//Proc. of the European Conference on Computer Vision, 2018: 3−19.
18	PARK J, WOO S, LEE J Y, et al. BAM: bottleneck attention modules[EB/OL]. [2025-04-21]. https://arxiv.org/abs/1807.06514.
19	王智睿, 康玉卓, 曾璇, 等. SAR-AIRcraft-1.0: 高分辨率SAR飞机检测识别数据集[J]. 雷达学报, 2023, 12 (4): 906- 922. doi: 10.12000/JR23043
	WANG Z R, KANG Y Z, ZENG X, et al. SAR-AIRcraft-1.0: high-resolution SAR aircraft detection and recognition dataset[J]. Journal of Radars, 2023, 12 (4): 906- 922. doi: 10.12000/JR23043
20	QIN R, FU X J, CHANG J Y, et al. Multilevel wavelet-SRNet for SAR target recognition[J]. IEEE Geoscience and Remote Sensing Letters, 2021, 19, 4009005.
21	CHO J H, PARK C G. Multiple feature aggregation using convolutional neural networks for SAR image-based automatic target recognition[J]. IEEE Geoscience and Remote Sensing Letters, 2018, 15 (12): 1882- 1886. doi: 10.1109/LGRS.2018.2865608
22	YANG J Y, KIM S. Noise robust SAR image classification using siamese spiking neural networks[J]. IEEE Geoscience and Remote Sensing Letters, 2024, 21, 4017205.
23	WANG X, WU Y X, SHI C T, et al. AEND-Net: adaptive noise estimation and despeckling network for SAR image[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024, 17, 4036- 4051. doi: 10.1109/JSTARS.2024.3355220
24	LI R, LIU M, CHEN S C, et al. LGM-RNet: large margin Gaussian mixture with ring loss network for imbalanced SAR images target recognition[J]. IEEE Geoscience and Remote Sensing Letters, 2024, 21, 4019005.
25	DING J, CHEN B, LIU H W, et al. Convolutional neural network with data augmentation for SAR target recognition[J]. IEEE Geoscience and Remote Sensing Letters, 2016, 13 (3): 364- 368.
26	BOUREAU Y L, PONCE J, LECUN Y. A theoretical analysis of feature pooling in visual recognition[C]//Proc. of the 27th International Conference on Machine Learning, 2010: 111−118.
27	KRIZHEVSKY A, SUTSKEVER I, HINTON G E. ImageNet classification with deep convolutional neural networks[J]. Communications of the ACM, 2017, 60 (6): 84- 90. doi: 10.1145/3065386
28	MOREIRA A. Improved multilook techniques applied to SAR and SCANSAR imagery[J]. IEEE Trans. on Geoscience and Remote Sensing, 1991, 29 (4): 529- 534. doi: 10.1109/36.135814
29	DAUBECHIES I. Ten lectures on wavelets[M]. Philadelphia: Society for Industrial and Applied Mathematics, 1992.
30	MALLAT S G. A theory for multiresolution signal decomposition: the wavelet representation[J]. IEEE Trans. on Pattern Analysis and Machine Intelligence, 1989, 11 (7): 674- 693. doi: 10.1109/34.192463
31	LI Q F, SHEN L L, GUO S, et al. Wavelet integrated CNNs for noise-robust image classification[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 7245−7254.
32	LI Q F, SHEN L L, GUO S, et al. WaveCNet: wavelet integrated CNNs to suppress aliasing effect for noise-robust image Classification[J]. IEEE Trans. on Image Processing, 2021, 30, 7074- 7089. doi: 10.1109/TIP.2021.3101395
33	YAO L Y, XU Y C, ZHANG S W, et al. HPA-UNet: a hybrid post-processing attention U-Net for tongue segmentation[J]. IEEE Journal of Biomedical and Health Informatics, 2025, 29(9): 6727−6739.
34	HE M J, ZHANG J, SHAN S G, et al. Locality-aware channel-wise dropout for occluded face recognition[J]. IEEE Trans. on Image Processing, 2021, 31, 788- 798.
35	HE D Y, GUO W W, ZHANG T, et al. Occluded target recognition in SAR imagery with scattering excitation learning and channel dropout[J]. IEEE Geoscience and Remote Sensing Letters, 2023, 20, 4005005.
36	YANG L, ZHANG R Y, LI L, et al. SimAM: a simple, parameter-free attention module for convolutional neural networks[C]//Proc. of the International Conference on Machine Learning, 2021: 11863−11874.
37	郭立民, 黄文青, 陈前, 等. 轻量化的ML-SNet雷达复合干扰识别算法[J]. 系统工程与电子技术, 2025, 47 (2): 418- 427.
	GUO L M, HUANG W Q, CHEN Q, et al. Lightweight algorithm of ML-SNet radar compound jamming recognition[J]. Systems Engineering and Electronics, 2025, 47 (2): 418- 427.
38	LEE J S. Digital image enhancement and noise filtering by use of local statistics[J]. IEEE Trans. on Pattern Analysis and Machine intelligence, 1980 (2): 165- 168.
39	FAWCETT T. An introduction to ROC analysis[J]. Pattern Recognition Letters, 2006, 27 (8): 861- 874. doi: 10.1016/j.patrec.2005.10.010

类别	训练集	测试集	总计
A220	278	31	309
A320/321	1 719	52	1 771
A330	3 270	460	3 730
ARJ21	825	362	1 187
Boeing737	2 007	550	2 557
Boeing787	2 191	454	2 645
Other	3 223	1 041	4 264

名称	参数
批量大小	30
优化器	Adam
初始学习率	0.0001
训练次数	100
权重衰减	0.0001

WSS模块	SimAM模块	准确率
×	×	0.8366
×	√	0.8576
√	×	0.8634
√	√	0.8701

模型	平均识别率	A220	A320/321	A330	ARJ21	Boeing737	Boeing787	Other
LCA-CNN	0.756 6	0.73	0.96	0.45	0.63	0.72	0.79	0.81
MFCNN	0.845 8	0.83	1.00	1.00	0.76	0.78	0.85	0.90
ResNet-101	0.830 8	0.84	0.98	1.00	0.88	0.81	0.81	0.82
ShuffleNetV2	0.835 9	0.80	1.00	1.00	0.82	0.80	0.81	0.87
Wavelet-SRNet	0.824 7	0.85	1.00	1.00	0.76	0.76	0.76	0.88
WSS-CNN	0.870 1	0.85	1.00	1.00	0.86	0.85	0.84	0.90

层类型	卷积核	输出通道数	参数量
Conv/BN/Relu/Maxpool	3×3	16	464
WSA	—	16	0
Conv/BN/Relu/Maxpool	3×3	32	4 672
Conv/BN/Relu/Avgpool	3×3	64	18 560
Conv/BN/Relu/Maxpool	3×3	128	73 980
Conv/BN/Relu	3×3	128	147 712
SimAM/Dropout	—	128	0
Conv/BN/Relu/Maxpool	3×3	256	295 424
FC	—	7	7 175
总计	—	—	547 991

基于小波散射激励与SimAM的轻量化抗噪卷积神经网络的目标识别

Target recognition of lightweight noise-resistant CNN based on wavelet scattering excitation and SimAM

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 17

参考文献 39

相关文章 15

编辑推荐

Metrics

本文评价

模型	参数量	显存占用/MB
LCA-CNN	2 557 374	10.288
MFCNN	49 612 031	197.79
ResNet-101	42 514 503	171.20
ShuffleNetV2	1 260 779	13.26
Wavelet-SRNet	960 685	11.86
WSS-CNN	547 991	10.42

方法	平均识别准确率
方法	无噪声	$a = 2$	$a = 5$
LCA-CNN	0.806 1	0.756 6	0.714 3
MFCNN	0.863 4	0.845 8	0.828 8
ResNet-101	0.839 6	0.830 8	0.796 1
ShuffleNetV2	0.846 7	0.835 9	0.806 1
Wavelet-SRNet	0.830 8	0.824 7	0.817 9
WSS-CNN	0.877 5	0.870 1	0.851 5

散斑噪声程度	梯度幅度值
散斑噪声程度	WSS+conv1	conv1
$a = 2$	2.5012	2.1019
$a = 5$	1.8508	1.4751

[1]	黄琼丹, 刘露露, 韩洁婧, 王佳鹏, 康仕林. 基于差异特征反投影融合的图像超分辨重建[J]. 系统工程与电子技术, 2026, 48(6): 1809-1818.
[2]	赵巍山, 尤思洋, 黄丽佳, 周光尧. 基于散射拓扑增强的双流SAR飞机检测网络[J]. 系统工程与电子技术, 2026, 48(6): 1848-1858.
[3]	袁永康, 王君旭, 华煜明, 张磊. 地基雷达对月多普勒补偿成像方法[J]. 系统工程与电子技术, 2026, 48(6): 1859-1868.
[4]	罗雨泉, 何雨强, 李雅鑫, 梁松, 王俊. 基于CNN-BiLSTM-MHA时空融合框架的毫米波雷达人体姿态估计[J]. 系统工程与电子技术, 2026, 48(5): 1502-1514.
[5]	孙景荣, 张华, 陈哲哲, 赵方正. 基于卷积神经网络的雾天道路能见度测量方法[J]. 系统工程与电子技术, 2026, 48(4): 1112-1124.
[6]	杨泽莹, 陈卓, 徐志明, 艾小锋, 吴其华, 王玲. 空间目标ISAR图像部件正向标记方法[J]. 系统工程与电子技术, 2026, 48(4): 1165-1173.
[7]	黄天佑, 林惠孚, 杨超, 赖涛, 王青松, 黄海风. SAR移频欺骗干扰距离主瓣展宽问题解决方法[J]. 系统工程与电子技术, 2026, 48(4): 1174-1185.
[8]	胡继军, 韩伟, 张国玉, 李洋, 杨洁, 田甜. 基于行为识别网络的SAR工作模式识别方法[J]. 系统工程与电子技术, 2026, 48(4): 1218-1229.
[9]	翟佳, 画麒嘉, 王梓权, 张子恺, 王睿琦, 刘金玲, 胡美琪, 武晨辉. 基于多级别融合的可见光和SAR图像联合识别[J]. 系统工程与电子技术, 2026, 48(3): 817-825.
[10]	李莹莹, 吴昊, 陈凤, 林郁卓越, 王雪莹, 冯鑫. 一种基于天线指向的大方位扫描滑聚模式星载SAR几何定位技术[J]. 系统工程与电子技术, 2026, 48(3): 846-858.
[11]	杨雪莹, 李高鹏, 张云. 语义引导的星载多角度SAR三维重建方法[J]. 系统工程与电子技术, 2026, 48(2): 430-446.
[12]	刘梦莉, 舒高峰, 李宁. 基于多干扰机协同调制的多通道SAR-GMTI欺骗干扰方法[J]. 系统工程与电子技术, 2026, 48(2): 490-502.
[13]	胡继军, 韩伟, 金龙, 周希娃, 张国玉, 周钦月. 雷达导引头的多站组网单比特干扰对抗方法[J]. 系统工程与电子技术, 2026, 48(2): 503-514.
[14]	陈海青, 汪刘应, 刘顾, 王龙, 葛超群, 陈孟州. 合成孔径雷达图像去噪算法研究进展[J]. 系统工程与电子技术, 2026, 48(1): 94-105.
[15]	张彬, 许高添, 张廷豪, 李志辉, 何宏强. 分布式合成孔径雷达前视高分辨成像算法[J]. 系统工程与电子技术, 2026, 48(1): 119-131.