基于CBAM-Swin-Transformer迁移学习的海上微动目标分类方法

doi:10.12305/j.issn.1001-506X.2025.04.12

系统工程与电子技术 ›› 2025, Vol. 47 ›› Issue (4): 1155-1167.doi: 10.12305/j.issn.1001-506X.2025.04.12

基于CBAM-Swin-Transformer迁移学习的海上微动目标分类方法

何肖阳¹, 陈小龙²^,*, 杜晓林¹, 苏宁远¹, 袁旺¹, 关键²

1. 烟台大学计算机与控制工程学院, 山东烟台 264005
2. 海军航空大学, 山东烟台 264001

收稿日期:2024-02-03 出版日期:2025-04-25 发布日期:2025-05-28
通讯作者: 陈小龙
作者简介:何肖阳 (1999—), 男, 硕士研究生, 主要研究方向为海杂波背景下目标检测
陈小龙 (1985—), 男, 教授, 博士, 主要研究方向为雷达低慢小目标检测、海杂波抑制、雷达智能信号处理
杜晓林 (1985—), 男, 副教授, 硕士研究生导师, 博士, 主要研究方向为雷达信号处理、波形设计、协方差矩阵估计
苏宁远 (1995—), 男, 博士研究生, 主要研究方向为雷达智能信号处理、海面目标检测
袁旺 (1997—), 男, 硕士研究生, 主要研究方向为高分辨雷达无人机等低慢小目标多特征融合识别
关键 (1968—), 男, 教授, 博士研究生导师, 博士, 主要研究方向为雷达目标检测与跟踪、侦察图像处理、信息融合
基金资助:
国家自然科学基金(62222120);山东省自然科学基金(ZR2024JQ003)

Classification of maritime micromotion target based on transfer learning in CBAM-Swin-Transformer

Xiaoyang HE¹, Xiaolong CHEN²^,*, Xiaolin DU¹, Ningyuan SU¹, Wang YUAN¹, Jian GUAN²

1. School of Computer and Control Engineering, Yantai University, Yantai 264005, China
2. Naval Aviation University, Yantai 264001, China

Received:2024-02-03 Online:2025-04-25 Published:2025-05-28
Contact: Xiaolong CHEN

摘要/Abstract

摘要：

雷达作为海上目标监测和识别的重要手段, 海上目标运动特征精细化描述与分类是其关键技术。基于深度学习的卷积网络分类方法不依赖于模型, 但仍难以适应复杂多变的海洋环境、多样性海上目标, 泛化能力有限。将卷积注意力机制模块(convolutional block attention module, CBAM)融入Swin-Transformer网络, 并基于迁移学习(transfer learning, TL)策略, 提出一种兼顾舰船目标和低空旋翼飞行目标的海上微动目标分类方法(简称为TL-CBAM-Swin-Transformer), 提升多种观测条件下的模型分类适应能力。首先, 建立海上微动目标模型, 并基于3种雷达实测数据构建海面非匀速平动、三轴转动、直升机、固定翼无人机的微动时频数据集。然后, 设计TL-CBAM-Swin-Transformer网络, CBAM从通道维和空间维提取特征, 提高其小尺度中多头注意力信息的提取能力。实测数据验证结果表明, 相比Swin-Transformer，所提网络的分类准确度提升3.43%。采用TL法, 将所提网络在ImageNet数据上进行预训练, 将智能像素处理(intelligent pixel processing, IPIX)雷达微动目标作为源域进行预训练，并迁移至科学与工业研究委员会(Council for Scientific and Industrial Research, CSIR)雷达微动目标, 分类概率达97.9%, 将直升机旋翼作为源域进行预训练并迁移至固定翼无人机, 分类概率达98.8%, 验证了所提算法具有较强的泛化能力。

关键词: 雷达目标分类, 海上微动目标, 迁移学习, Swin-Transformer网络, 注意力机制, 时频分析

Abstract:

As an important means for maritime target detection and identification, radar requires fine-grained description and classification of the motion characteristics of maritime targets, which is a key technology. Deep learning-based convolutional network classification method, although model-independent, still struggle to adapt to the complex and diverse maritime environment and the variety of maritime targets, with limited generalization ability. Convolutional block attention module (CBAM) is integrated into the Swin-Transformer network, and based on transfer learning (TL) strategy, a maritime micromotion target classification method (TL-CBAM-Swin-Transformer) is proposed to consider both ship targets and low-altitude rotorcraft flight targets, thereby enhancing the model classification adaptability under various observation conditions. Firstly, a maritime micromotion target model is established, and based on three radar measurement data sets, micromotion time-frequency datasets of sea surface non-uniform translational motion, three-axis rotation, helicopters, and fixed rotor drones are constructed. Then, the TL-CBAM-Swin-Transformer network is designed, where CBAM extracts features from both the channel and spatial dimensions, enhancing its ability to extract multi-head attention information at small scales. Experimental data verification result shows that compared to Swin-Transformer, the classification accuracy of the proposed algorithm is improved by 3.43%. By using TL method, the proposed network is pre-trained on the ImageNet dataset and is transferred to Council for Scientific and Industrial Research (CSIR) micromotion targets with intelligent pixel processing (IPIX) radar micromotion targets as the source domain for pre-training, achieving a classification probability of 97.9%. With helicopter rotors as the source domain for pre-training, the proposed algorithm transfer to fixed rotorcraft drones, achieving a classification probability of 98.8%, which vadidates the strong generalization ability of the proposed algorithm.

Key words: radar target classification, maritime micromotion target, transfer learning (TL), Swin-Transformer network, attention mechanism, time-frequency analysis

中图分类号:

TN958.2

何肖阳, 陈小龙, 杜晓林, 苏宁远, 袁旺, 关键. 基于CBAM-Swin-Transformer迁移学习的海上微动目标分类方法[J]. 系统工程与电子技术, 2025, 47(4): 1155-1167.

Xiaoyang HE, Xiaolong CHEN, Xiaolin DU, Ningyuan SU, Wang YUAN, Jian GUAN. Classification of maritime micromotion target based on transfer learning in CBAM-Swin-Transformer[J]. Systems Engineering and Electronics, 2025, 47(4): 1155-1167.

图/表 24

图1

图2

图3

图4

图5

图6

表1

图7

图8

表2

图9

图10

表3

图11

图12

图13

图14

图15

图16

表4

图17

图18

图19

图20

参考文献 32

12	WANG X , ZHOU Y P , ZHOU D Q , et al. Research on low probability of intercept radar signal recognition using deep belief network and bispectra diagonal slice[J]. Journal of Electronics & Information Technology, 2016, 38 (11): 2972- 2976.
13	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
14	DENG J, DONG W, SOCHER R, et al. ImageNet: a large-scale hierarchical image database[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2009: 248-255.
15	苏宁远, 陈小龙, 关键, 等. 基于卷积神经网络的海上微动目标检测与分类方法[J]. 雷达学报, 2018, 7 (5): 565- 574.
	SU N Y , CHEN X L , GUAN J , et al. Detection and classification of maritime target with micro-motion based on CNNs[J]. Journal of Radars, 2018, 7 (5): 565- 574.
16	WANG J G , LI S B . Maritime radar target detection in sea clutter based on CNN with dual-perspective attention[J]. IEEE Geoscience and Remote Sensing Letters, 2022, 20, 3500405.
17	MENG X W . Rank sum nonparametric CFAR detector in nonhomogeneous background[J]. IEEE Trans. on Aerospace and Electronic Systems, 2021, 57 (1): 397- 403.
18	李鑫, 陆伟, 马召祎, 等. 基于图注意力和改进Transformer的节点分类方法[J]. 电子学报, 2024, 52 (8): 2799- 2810.
	LI X , LU W , MA Z Y , et al. A node classification method based on graph attention and improved transformer[J]. Acta Electronica Sinica, 2024, 52 (8): 2799- 2810.
19	LIU Z, LIN Y T, CAO Y, et al. Swin transformer: hierarchical vision transformer using shifted windows[C]//Proc. of the IEEE/CVF International Conference on Computer Vision, 2021: 10012-10022.
20	LI Y . Review of application of transfer learning in medical image analysis[J]. Computer Engineering and Applications, 2021, 57 (20): 42- 52.
21	张雪松, 庄严, 闫飞, 等. 基于迁移学习的类别级物体识别与检测研究与进展[J]. 自动化学报, 2019, 45 (7): 1224- 1243.
	ZHANG X S , ZHUANG Y , YAN F , et al. Status and deve-lopment of transfer learning based category-level object recognition and detection[J]. Acta Automatica Sinica, 2019, 45 (7): 1224- 1243.
22	ZHOU C C , ZHANG L K , ZENG G P , et al. Dispersed computing resource discovery model and algorithm for polymorphic migration network architecture[J]. Chinese Journal of Electronics, 2023, 32 (4): 821- 839.
23	QU Y Z , LIU W J , WANG J X , et al. Enhanced CNN-based small target detection in sea clutter with controllable false alarm[J]. IEEE Sensors Journal, 2023, 23 (9): 10193- 10205.
24	GUO C, WANG H P, XIA X Z, et al. Deep transfer learning based method for radar automatic recognition with small data size[C]//Proc. of the IEEE International Conference on Unmanned Systems, 2021: 995-999.
25	任硕良, 索继东, 佟禹. 卷积神经网络结合迁移学习的SAR目标识别[J]. 电光与控制, 2020, 27 (10): 37- 41.
	REN S L , SUO J D , TONG Y . SAR target recognition based on convolution neural network and transfer learning[J]. Electronics Optics & Control, 2020, 27 (10): 37- 41.
26	陈小龙, 董云龙, 李秀友, 等. 海面刚体目标微动特征建模及特性分析[J]. 雷达学报, 2015, 4 (6): 630- 638.
	CHEN X L , DONG Y L , LI X Y , et al. Modeling of micromotion and analysis of properties of rigid marine targets[J]. Journal of Radars, 2015, 4 (6): 630- 638.
27	李宇倩, 易建新, 万显荣, 等. 外辐射源雷达直升机旋翼参数估计方法[J]. 雷达学报, 2018, 7 (3): 313- 319.
	LI Y Q , YI J X , WAN X R , et al. Helicopter rotor parameter estimation method for passive radar[J]. Journal of Radars, 2018, 7 (3): 313- 319.
1	XU S W , ZHENG J B , PU J , et al. Sea-surface floating small target detection based on polarization features[J]. IEEE Geoscience and Remote Sensing Letters, 2018, 15 (10): 1505- 1509.
2	陈小龙, 关键, 何友. 微多普勒理论在海面目标检测中的应用及展望[J]. 雷达学报, 2013, 2 (1): 123- 134.
	CHEN X L , GUAN J , HE Y . Applications and prospect of micro-motion theory in the detection of sea surface target[J]. Journal of Radars, 2013, 2 (1): 123- 134.
3	罗迎, 张群, 王国正, 等. 基于复图像OMP分解的宽带雷达微动特征提取方法[J]. 雷达学报, 2012, 1 (4): 361- 369.
	LUO Y , ZHANG Q , WANG G Z , et al. Micro motion signature extraction method for wideband radar based on complex image OMP decomposition[J]. Journal of Radars, 2012, 1 (4): 361- 369.
4	LI Y M , WU R B , BAO Z , et al. Inverse synthetic aperture radar imaging of ship target with complex motion[J]. IET Radar, Sonar & Navigation, 2008, 2 (6): 395- 403.
5	HAJDUCH G , LE-CAILLEC J M , GARELLO R . Airborne high-resolution ISAR imaging of ship targets at sea[J]. IEEE Trans. on Aerospace and Electronic Systems, 2004, 40 (1): 378- 384. doi: 10.1109/TAES.2004.1292177
6	CHEN X L , GUAN J , BAO Z H , et al. Detection and extraction of target with micromotion in spiky sea clutter via short-time fractional Fourier transform[J]. IEEE Trans. on Geoscience and Remote Sensing, 2014, 52 (2): 1002- 1018. doi: 10.1109/TGRS.2013.2246574
7	CHEN X L , GUAN J , LI X Y , et al. Effective coherent integration method for marine target with micromotion via phase differentiation and radon- Lv's distribution[J]. IET Radar, Sonar & Navigation, 2015, 9 (9): 1284- 1295.
8	WAGNER S A . SAR ATR by a combination of convolutional neural network and support vector machines[J]. IEEE Trans. on Aerospace and Electronic Systems, 2016, 52 (6): 2861- 2872. doi: 10.1109/TAES.2016.160061
9	KIM Y , TOOMAJIAN B . Hand gesture recognition using micro-doppler signatures with convolutional neural network[J]. IEEE Access, 2016, 4, 7125- 7130. doi: 10.1109/ACCESS.2016.2617282
10	王俊, 郑彤, 雷鹏, 等. 深度学习在雷达中的研究综述[J]. 雷达学报, 2018, 7 (4): 395- 411.
	WANG J , ZHENG T , LEI P , et al. Survey of study on deep learning in radar[J]. Journal of Radars, 2018, 7 (4): 395- 411.
11	徐彬, 陈渤, 刘宏伟, 等. 基于注意循环神经网络模型的雷达高分辨率距离像目标识别[J]. 电子与信息学报, 2016, 38 (12): 2988- 2995.
	XU B , CHEN B , LIU H W , et al. Attention-based recurrent neural network model for radar high-resolution range profile target recognition[J]. Journal of Electronics & Information Technology, 2016, 38 (12): 2988- 2995.
12	王星, 周一鹏, 周冬青, 等. 基于深度置信网络和双谱对角切片的低截获概率雷达信号识别[J]. 电子与信息学报, 2016, 38 (11): 2972- 2976.
28	万磊, 佟鑫, 盛明伟, 等. Softmax分类器深度学习图像分类方法应用综述[J]. 导航与控制, 2019, 18 (6): 1-9, 47.
	WAN L , TONG X , SHENG M W , et al. Review of image classification based on Softmax classifier in deep learning[J]. Navigation and Control, 2019, 18 (6): 1-9, 47.
29	陈小龙, 关键, 何友, 等. 高分辨稀疏表示及其在雷达动目标检测中的应用[J]. 雷达学报, 2017, 6 (3): 239- 251.
	CHEN X L , GUAN J , HE Y , et al. High-resolution sparse representation and its applications in radar moving target detection[J]. Journal of Radars, 2017, 6 (3): 239- 251.
30	陈小龙, 袁旺, 杜晓林, 等. 多波段FMCW雷达低慢小探测数据集(LSS-FMCWR-1.0)及高分辨微动特征提取方法[J]. 雷达学报(中英文版), 2024, 13 (2): 539- 553.
	CHEN X L , YUAN W , DU X L , et al. Multiband FMCW radar LSS-target detectiondataset (LSS-FMCWR-1.0) and high-resolution micromotion feature extraction method[J]. Journal of Radars, 2024, 13 (3): 539- 553.
31	李振春, 刁瑞, 韩文功, 等. 线性时频分析方法综述[J]. 勘探地球物理进展, 2010, 33 (4): 239- 246.
	LI Z C , DIAO R , HAN W G , et al. Review on linear time-frequency analysis methods[J]. Progress in Exploration Geophysics, 2010, 33 (4): 239- 246.
32	胡松, 江小炜, 杨光, 等. 滑动平均滤波在微弱脉冲信号检测中的应用[J]. 计算机与数字工程, 2007, 35 (10): 169- 171.
	HU S , JIANG X W , YANG G , et al. Using of moving average filter in faint pulse signal detection[J]. Computer and Digital Engineering, 2007, 35 (10): 169- 171.

数据集种类	雷达类型	数据类型	训练集/张	测试集/张	目标种类	信噪比/dB
IPIX	X波段IPIX雷达^[15]	实测海杂波+仿真目标	25 981	10 400	匀变速目标、变加速目标、微动Ⅰ、微动Ⅱ	[-30, 10]
CSIR	X波段CSIR雷达^[29]	实测数据	-	1 000	匀变速目标、变加速目标、微动Ⅰ、微动Ⅱ	-
直升机	X波段雷达^[27]	实测海杂波+仿真目标	1 200	1 040	单翼机、双翼机、多翼机	[-20, 10]
无人机	Ku波段雷达^[29]	实测数据	-	125	固定翼无人机	[-5, 5]

运动类型	初速度/(m/s)	加速度/(m/s²)	变加速度/(m/s³)	角速度/(rad/s)	微动周期/s	采样点数
匀加速	[-2, 10]	[-5, 5]	-	-	-	2¹⁰Hz, 1 s
非匀变速	[-5, 15]	[-5, 5]	[-8, 8]	-	-	2¹¹ Hz, 0.5 s
微动Ⅰ	-	-	-	ω_x=[0.3, 0.38] ω_y=[0.1, 0.15] ω_z=[0.06, 0.08]	T_x=26.4 T_y=11.2 T_z=33.2	2¹¹ Hz, 0.5 s
微动Ⅱ	-	-	-	ω_x=[0.5, 0.58] ω_y=[0.8, 0.95] ω_z=[0.54, 0.58]	T_x=12.2 T_y=6.8 T_z=14.2	2¹¹Hz, 0.5 s

参数	取值
距离分辨力/m	0.5
观测时间/s	1
采样点数/个	10 240
叶片长度和宽度/m	L=6, W=1
旋转速率/(rad/s)	4

模型	分类准确率/%	测试用时/s
TL-CBAM-Swin-Transformer	98.63	5.94
Swin-Transformer	95.20	14.60
LeNet	93.43	6.50
AlexNet	82.16	8.92
GoogLeNet	89.60	10.23

[1]	何肖阳, 陈小龙, 杜晓林, 苏宁远, 袁旺, 关键. 基于CBAM-Swin-Transformer迁移学习的海上微动目标分类方法[J]. 系统工程与电子技术, 2025, 47(4): 1155-1167.
[2]	聂千祁, 沙明辉, 朱应申, 王崇宇, 崔念强. 基于盲源分离结合奇异谱分析的雷达多分量信号识别方法[J]. 系统工程与电子技术, 2025, 47(4): 1168-1175.
[3]	李家宽, 冯博, 刘红亮, 叶春茂, 余继周. 基于角度引导注意力的气动目标宽带PD识别方法[J]. 系统工程与电子技术, 2025, 47(3): 807-816.
[4]	冯蕴雯, 潘维煌, 路成, 刘佳奇. 基于故障逻辑的民机液压状态监控与故障诊断[J]. 系统工程与电子技术, 2025, 47(3): 842-854.
[5]	宋怡佳, 王海燕, 冯伟, 全英汇. 基于高光谱跨场景迁移学习的集成对齐子空间自适应算法[J]. 系统工程与电子技术, 2025, 47(2): 369-375.
[6]	安强, 叶春茂, 鲁耀兵, 张彦. 复杂场景下间歇采样转发干扰抑制算法[J]. 系统工程与电子技术, 2025, 47(2): 463-477.
[7]	刘强, 孙浩冉, 胡邓华, 张爽. 基于Vondrak-Cepek组合滤波和注意力机制加权的时间比对融合算法[J]. 系统工程与电子技术, 2025, 47(2): 673-679.
[8]	陈晓萱, 徐书文, 胡绍海, 马晓乐. 基于卷积与自注意力的红外与可见光图像融合[J]. 系统工程与电子技术, 2024, 46(8): 2641-2649.
[9]	张瑞斌, 朱梦韬, 李云杰. 对调制识别网络隐身的雷达发射信号生成方法[J]. 系统工程与电子技术, 2024, 46(7): 2256-2268.
[10]	华煜明, 王东亚, 朱天林, 金胜, 王洋. 复合微多普勒信号的时频谐波分析方法[J]. 系统工程与电子技术, 2024, 46(7): 2301-2309.
[11]	吴嘉俊, 苏春, 张玉茹. 基于双重自注意力机制和长短时记忆网络的剩余寿命预测[J]. 系统工程与电子技术, 2024, 46(6): 1986-1994.
[12]	薛丽莎, 葛瑞星, 朱宇轩, 鲍雁飞. 基于SABEGAN的通信干扰信号生成与效能分析[J]. 系统工程与电子技术, 2024, 46(6): 2155-2163.
[13]	牛闯, 林强, 段敏, 施端阳, 谷成刚. 脉内频率-时延捷变雷达抗间歇采样转发干扰方法[J]. 系统工程与电子技术, 2024, 46(5): 1583-1598.
[14]	刘绍华, 杜康, 佘春东, 杨傲. 基于CenterNet的多教师联合知识蒸馏[J]. 系统工程与电子技术, 2024, 46(4): 1174-1184.
[15]	方小宇, 黄丽佳. 基于全局位置信息和残差特征融合的SAR船舶检测算法[J]. 系统工程与电子技术, 2024, 46(3): 839-848.

基于CBAM-Swin-Transformer迁移学习的海上微动目标分类方法

Classification of maritime micromotion target based on transfer learning in CBAM-Swin-Transformer

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 24

参考文献 32

相关文章 15

编辑推荐

Metrics

本文评价