基于EfficientNet的多通道雷达目标微动特征分类方法

doi:10.12305/j.issn.1001-506X.2024.09.11

摘要/Abstract

摘要：

针对在低空雷达监视场景下, 行人、车辆、无人机等目标分类任务中目标微动特征难以提取导致分类准确率较低的问题, 提出一种基于EfficientNet的多通道雷达目标微动特征分类方法。首先, 根据杂波、目标和噪声信号的能量分布差异, 提出多能量奇异值分解方法抑制杂波和噪声, 增强目标微动特征。随后, 联合雷达和差通道时频信息特点, 设计多通道EfficientNet模型, 结合多通道微动特性进一步实现目标的准确分类。最后, 利用雷达实测目标数据对所提方法进行验证。结果表明, 所提方法在保证较低模型复杂度的情况下, 相比于其他方法在准确率上有显著提升。

关键词: 多通道雷达, 微动目标分类, EfficientNet, 奇异值分解

Abstract:

To solve the problem on low classification accuracy caused by the difficulty of extracting features of micro-motion targets, such as pedestrians, vehicles, and drones in low-altitude radar monitoring scenarios, a multi-channel radar target micro-motion feature classification method based on EfficientNet is proposed. Firstly, a multi-energy singular value decomposition method is proposed to suppress clutter and noise, and enhance the micro-motion characteristics of the target based on the energy distribution differences between the clutter, target, and noise signals. Secondly, the multi-channel EfficientNet model is designed to combine time-frequency information features in radar sum and difference channels, and further fuse multi-channel micro-motion features to achieve accurate target classification. Finally, the effectiveness of the proposed method is verified through experiments using radar measured target data. The results show that compared with other methods, the proposed method significantly improves classification accuracy with low model complexity.

Key words: multi-channel radar, classification of micro-motion targets, EfficientNet, singular value decomposition

中图分类号:

TN959.1

王潇怡, 罗运华, 喻忠军, 孙浩, 王晓蓓. 基于EfficientNet的多通道雷达目标微动特征分类方法[J]. 系统工程与电子技术, 2024, 46(9): 2988-2998.

Xiaoyi WANG, Yunhua LUO, Zhongjun YU, Hao SUN, Xiaobei WANG. Multi-channel radar target micro-motion feature classification method based on EfficientNet[J]. Systems Engineering and Electronics, 2024, 46(9): 2988-2998.

图/表 20

图1

图2

图3

图4

图5

表1

图6

图7

图8

图9

图10

图11

图12

图13

表2

图14

表3

图15

表4

表5

参考文献 33

9	AMIN M G, ZENG Z X, SHAN T, et al. Hand gesture recognition based on radar micro-Doppler signature envelopes[C]//Proc. of the IEEE Radar Conference, 2019.
10	HANIF A , MUAZ M , HASAN A , et al. Micro-Doppler based target recognition with radars: a review[J]. IEEE Sensors Journal, 2022, 22 (4): 2948- 2961. doi: 10.1109/JSEN.2022.3141213
11	杨磊, 孙卫天, 毛欣瑶, 等. 雷达鸟类目标微多普勒贝叶斯增强算法[J]. 系统工程与电子技术, 2024, 46 (2): 505- 516.
	YANG L , SUN W T , MAO X Y , et al. Bayesian enhancement algorithm for micro-Doppler feature of radar bird target[J]. Systems Engineering and Electronics, 2024, 46 (2): 505- 516.
12	JI H R , HOU C P , YANG Y , et al. A one-class classification method for human gait authentication using micro-Doppler signatures[J]. IEEE Signal Processing Letters, 2021, 28, 2182- 2186. doi: 10.1109/LSP.2021.3122344
13	KANG H S , KIM B K , PARK J S , et al. Drone elevation angle classification based on convolutional neural network with micro-Doppler of multi-polarization[J]. IEEE Geoscience and Remote Sensing Letters, 2022, 19, 3501205.
14	ZHU L Z , ZHANG S N , CHEN K Y , et al. Low-SNR recognition of UAV-to-ground targets based on micro-Doppler signatures using deep convolutional denoising encoders and deep residual learning[J]. IEEE Trans.on Geoscience and Remote Sensing, 2022, 60, 5106913.
15	HE W K , SUN J B , ZHANG X Y , et al. Micro-Doppler feature extraction of micro-rotor UAV under the background of low SNR[J]. Journal of Systems Engineering and Electronics, 2022, 33 (6): 1127- 1139.
16	ZHAO X Z , YE B Y . Selection of effective singular values using difference spectrum and its application to fault diagnosis of headstock[J]. Mechanical Systems and Signal Processing, 2011, 25 (5): 1617- 1631. doi: 10.1016/j.ymssp.2011.01.003
17	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
18	查翔, 倪世宏, 张鹏. 一类非线性信号去噪的奇异值分解有效迭代方法[J]. 电子与信息学报, 2015, 37 (6): 1330- 1335.
	ZHA X , NI S H , ZHANG P . Effective iteration method of a class of nonlinear signal denoising based on singular value decomposition[J]. Journal of Electronics & Information Techno-logy, 2015, 37 (6): 1330- 1335.
19	夏朝阳, 周成龙, 介钧誉, 等. 基于多通道调频连续波毫米波雷达的微动手势识别[J]. 电子与信息学报, 2020, 42 (1): 164- 172.
	XIA Z Y , ZHOU C L , JIE J Y , et al. Micro-motion gesture recognition based on multi-channel frequency modulated continuous wave millimeter wave radar[J]. Journal of Electronics & Information Technology, 2020, 42 (1): 164- 172.
20	李浩冬, 廖桂生, 许京伟. 弹载雷达和差通道稳健自适应杂波抑制方法[J]. 系统工程与电子技术, 2019, 41 (2): 273- 279.
	LI H D , LIAO G S , XU J W . Robust adaptive clutter suppression approach for missile-borne radar with ∑Δ-beam[J]. Systems Engineering and Electronics, 2019, 41 (2): 273- 279.
21	MA X J , QI W , CHE K J , et al. Parameter estimation of LFM signals based on time reversal[J]. Journal of Systems Engineering and Electronics, 2023, 34 (3): 674- 681. doi: 10.23919/JSEE.2023.000014
22	ABRATKIEWICZ K , SAMCZYNSKI P J , RYTEL-ANDRIANIK R , et al. Multipath interference removal in receivers of linear frequency modulated radar pulses[J]. IEEE Sensors Journal, 2021, 21 (17): 19000- 19012. doi: 10.1109/JSEN.2021.3087319
23	ZHANG L , JIANG H D , ZHOU Y R , et al. Wheeled vehicle detection under the asymmetric micro-Doppler interference[J]. IEEE Sensors Journal, 2022, 22 (18): 18085- 18092. doi: 10.1109/JSEN.2022.3196317
24	KARABACAK C , GURBUZ S Z , GURBUZ A C , et al. Know-ledge exploitation for human micro-Doppler classification[J]. IEEE Geoscience and Remote Sensing Letters, 2015, 12 (10): 2125- 2129. doi: 10.1109/LGRS.2015.2452311
25	王凡, 刘丽, 徐航, 等. 基于自适应阈值滤波和S-Method的穿墙人体动作识别[J]. 电子器件, 2021, 44 (5): 1265- 1273.
1	GAO Y , ZHOU Y , WANG Y , et al. Narrowband radar automatic target recognition based on a hierarchical fusing network with multi-domain features[J]. IEEE Geoscience and Remote Sensing Letters, 2021, 18 (6): 1039- 1043. doi: 10.1109/LGRS.2020.2993039
2	CAO X M , YI J X , GONG Z P , et al. Automatic target recognition based on RCS and angular diversity for multi-static passive radar[J]. IEEE Trans.on Aerospace and Electronic Systems, 2022, 58 (5): 4226- 4240. doi: 10.1109/TAES.2022.3159295
3	YU X L, LIU S, REN H H, et al. Transductive prototypical attention network for few-shot SAR target recognition[C]//Proc. of the IEEE Radar Conference, 2023.
4	CHEN V C , LI F , HO S S , et al. Micro-Doppler effect in radar: phenomenon, model, and simulation study[J]. IEEE Trans.on Aerospace and Electronic Systems, 2006, 42 (1): 2- 21.
5	李彦兵, 杜兰, 刘宏伟, 等. 基于微多普勒效应和多级小波分解的轮式履带式车辆分类研究[J]. 电子与信息学报, 2013, 35 (4): 894- 900.
	LI Y B , DU L , LIU H W , et al. Study on classification of wheeled and tracked vehicles based on micro-Doppler effect and multilevel wavelet decomposition[J]. Journal of Electronics & Information Technology, 2013, 35 (4): 894- 900.
6	王福友, 罗钉, 刘宏伟. 低分辨机载雷达飞机目标分类识别技术研究[J]. 雷达学报, 2014, 3 (4): 444- 449.
	WANG F Y , LUO D , LIU H W , et al. Low-resolution airborne radar aircraft target classification[J]. Journal of Radars, 2014, 3 (4): 444- 449.
7	LI Y B , DU L , LIU H W . Hierarchical classification of moving vehicles based on empirical mode decomposition of micro-Doppler signatures[J]. IEEE Trans.on Geoscience and Remote Sensing, 2013, 51 (5): 3001- 3013. doi: 10.1109/TGRS.2012.2216885
8	SVENNINGSSON P , FIORANELLI F , YAROVOY A , et al. A Bayesian network for the classification of human motion as observed by distributed radar[J]. IEEE Trans.on Aerospace and Electronic Systems, 2022, 58 (6): 5661- 5674. doi: 10.1109/TAES.2022.3177589
25	WANG F , LIU L , XU H , et al. Through-wall human motion classification based on adaptive threshold filtering and S-Method[J]. Chinese Journal of Electron Devices, 2021, 44 (5): 1265- 1273.
26	JOSHI S K , BAUMGARTNER S V , CARDOSO D S A B , et al. Direction-of-arrival angle and position estimation for extended targets using multichannel airborne radar data[J]. IEEE Geoscience and Remote Sensing Letters, 2022, 19, 4022705.
27	ROCCA P , MORABITO A F . Optimal synthesis of reconfigurable planar arrays with simplified architectures for monopulse radar applications[J]. IEEE Trans.on Antennas and Propagation, 2015, 63 (3): 1048- 1058. doi: 10.1109/TAP.2014.2386359
28	TAN M, LE Q V. EfficientNet: rethinking model scaling for convolutional neural networks[C]//Proc. of the 36th International Conference on Machine Learning, 2019.
29	HU J, SHEN L, ALBANIE S, et al. Squeeze-and-excitation networks[C]//Proc. of the 31st IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018: 7132-7141.
30	SAVARE G , TOSCANI G . The concavity of Renyi entropy power[J]. IEEE Trans.on Information Theory, 2014, 60 (5): 2687- 2693. doi: 10.1109/TIT.2014.2309341
31	WONG T T , YANG N Y . Dependency analysis of accuracy estimates in k-fold cross validation[J]. IEEE Trans.on Know-ledge and Data Engineering, 2017, 29 (11): 2417- 2427. doi: 10.1109/TKDE.2017.2740926
32	SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[EB/OL]. [2023-04-12]. http://arxiv.org/pdf/1409.1556.pdf.
33	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770-778.

阶段	模块	输出尺寸	层数
1	Conv3×3	56×56×32	1
2	MBConv1, k3×3	56×56×16	1
3	MBConv6, k5×5	28×28×24	2
4	MBConv6, k5×5	14×14×40	2
5	MBConv6, k3×3	7×7×80	3
6	MBConv6, k5×5	7×7×112	3
7	MBConv6, k5×5	3×3×192	4
8	MBConv6, k3×3	3×3×320	1
9	Conv1×1&Pooling&FC	1×k	1

名称	设置参数
批量大小	16
优化器	随机梯度下降
优化器学习率	0.01
损失函数	交叉熵损失函数

实验	分类方法	特征	通道数	准确率/%
1	SVM	时域	1	74.19
2	MESVD-SM-SVM	时域	1	78.26
3	MESVD-SM-SVM	频域	1	74.46
4	MESVD-SM-SVM	时频域	1	78.26
5	MESVD-SM-SVM	多域	1	80.43
6	MESVD-SM-SVM	多域	2	83.70
7	STFT-EfficientNet	微多普勒	1	96.52
8	MESVD-SM-EfficientNet	微多普勒	1	97.61
9	MESVD-SM-EfficientNet	微多普勒	2	99.13

序号	尺寸	通道	准确率/%	FLOPs(×10⁹)/次
1	56×56	1	93.70	0.03
2	112×112	1	97.61	0.11
3	224×224	1	97.61	0.41
4	56×56	2	96.96	0.07
5	112×112	2	99.13	0.23
6	224×224	2	98.91	0.83

序号	网络模型	准确率/%	参数量(×10⁶)	FLOPs(×10⁹)/次
1	AlexNet	91.09	61.60	0.41
2	VggNet19	89.13	144.20	10.0
3	Resnet50	98.70	26.10	2.18
4	EfficientNet	99.13	5.80	0.23