HRRP稀疏自编码器深层特征与散射中心特征的关联性研究

doi:10.12305/j.issn.1001-506X.2021.11.02

系统工程与电子技术 ›› 2021, Vol. 43 ›› Issue (11): 3040-3053.doi: 10.12305/j.issn.1001-506X.2021.11.02

HRRP稀疏自编码器深层特征与散射中心特征的关联性研究

霍超颖*, 闫华, 冯雪健, 殷红成, 邢笑宇, 陆金文

北京环境特性研究所电磁散射重点实验室, 北京 100854

收稿日期:2021-02-23 出版日期:2021-11-01 发布日期:2021-11-12
通讯作者: 霍超颖
作者简介:霍超颖(1982—), 女, 研究员, 博士研究生, 主要研究方向为雷达目标散射特性、电磁散射特征提取与应用|闫华(1981—), 男, 高级工程师, 博士, 主要研究方向为雷达目标散射特性、计算电磁学、特征提取、参数化建模|冯雪健(1991—), 男, 工程师, 博士, 主要研究方向为雷达目标散射特性、目标电磁逆散射、电磁参数反演|殷红成(1967—), 男, 研究员, 博士, 主要研究方向为雷达目标特性、计算电磁学、目标识别|邢笑宇(1987—), 女, 高级工程师, 博士研究生, 主要研究方向为雷达目标散射特性、电磁散射参数化建模、电磁散射特征提取与识别|陆金文(1994—), 男, 博士研究生, 主要研究方向为雷达目标散射特性、电磁散射参数化建模
基金资助:
国家重点研发计划(2018YFC0825804)

Correlation research between deep features of HRRP sparse auto-encoder and scattering center features

Chaoying HUO*, Hua YAN, Xuejian FENG, Hongcheng YIN, Xiaoyu XING, Jinwen LU

Science and Technology on Electromagnetic Scattering Laboratory, Beijing Institute ofEnvironmental Features, Beijing 100854, China

Received:2021-02-23 Online:2021-11-01 Published:2021-11-12
Contact: Chaoying HUO

摘要/Abstract

摘要：

利用稀疏自编码器网络对典型目标一维高分辨距离像(high resolution range profile, HRRP)进行了学习训练, 基于各层权重系数矩阵定义了一种综合权重系数, 通过综合权重系数和降维特征与散射中心特征的对比分析, 发现稀疏自编码器深层特征与散射中心特征之间具有一定的关联性, 并对综合权重系数和深层降维特征的物理意义进行了解释。首先针对HRRP构建稀疏自编码器网络, 经过深层学习后获取训练后的权重系数和降维后的特征, 并与散射中心的位置特征和强度分布特征进行关联性分析。结果表明, 综合权重系数矩阵为与散射中心密切相关的类字典系数矩阵, 反映了距离域强散射中心位置随角度变化的可能的分子集; 降维特征能够实现对强散射中心的学习和提取, 反映了强散射中心位置和强度随角度的变化。最后分析了网络训练层数和降维维数对学习训练结果的影响, 可指导后续网络参数的选择。文章首次针对雷达HRRP数据开展深度学习特征的可解释性研究, 为后续深度学习在雷达数据处理中的广泛应用提供了有益的导引。

关键词: 稀疏自编码器, 权重系数, 降维特征, 散射中心特征

Abstract:

A sparse auto-encoder network is used to learn and train one-dimensional high resolution range profile (HRRP) of typical targets. A comprehensive weight coefficient is defined based on the weight coefficient matrix of each layer. By comparing the weight coefficient and dimension reduction feature with the scattering center feature, it is found that there is a certain correlation between the deep feature of sparse auto-encoder and the scattering center feature. And the physical meaning of the comprehensive weight coefficient and deep dimension reduction feature is explained in this paper. Firstly, a sparse auto-encoder network is constructed for HRRP. After deep learning, the weight coefficient after training and the feature after dimension reduction are obtained, and the correlation with the position feature and intensity distribution feature of the scattering center is studied. The results show that the comprehensive weight coefficient matrix is a dictionary-like coefficient matrix closely related to the scattering center, which reflects the possible molecule set of the strong scattering center position changing with the angle in the range domain; and the dimension reduction feature can realize the learning and extraction of the strong scattering center, which reflects the change of the strong scattering center position and intensity with the angle. Finally, the influence of the number of training layers and the dimension reduction dimension on the learning and training results is analyzed, which can guide the selection of the subsequent network parameters. For the first time, this paper studies the interpretability of deep learning features for radar HRRP data, which provides a useful guidance for the subsequent extensive application of deep learning in radar data processing.

Key words: high resolution range profile (HRRP), sparse auto-encoder, deep feature, scattering center feature

中图分类号:

TN95

霍超颖, 闫华, 冯雪健, 殷红成, 邢笑宇, 陆金文. HRRP稀疏自编码器深层特征与散射中心特征的关联性研究[J]. 系统工程与电子技术, 2021, 43(11): 3040-3053.

Chaoying HUO, Hua YAN, Xuejian FENG, Hongcheng YIN, Xiaoyu XING, Jinwen LU. Correlation research between deep features of HRRP sparse auto-encoder and scattering center features[J]. Systems Engineering and Electronics, 2021, 43(11): 3040-3053.

图/表 22

图1

图2

图3

表1

图4

图5

图6

图7

图8

图9

图10

图11

图12

图13

图14

图15

图16

图17

表2

图18

图19

图20

参考文献 21

1	袁莉. 基于高分辨距离像的雷达目标识别方法研究[D]. 西安: 西安电子科技大学, 2007.
	YUAN L. Study of radar target recognition method base on high range resolution profile[D]. Xi'an: Xidian University, 2007.
2	高倩, 刘家学, 吴仁彪. 基于平均模板的HRRP自动目标识别[J]. 现代雷达, 2003, 25 (3): 14- 17. doi: 10.3969/j.issn.1004-7859.2003.03.005
	GAO Q , LIU J X , WU R B . Automatic target recognition based on HRRP mean-templates[J]. Modern Radar, 2003, 25 (3): 14- 17. doi: 10.3969/j.issn.1004-7859.2003.03.005
3	高倩, 吴仁彪. 一种基于SVM分类器的HRRP-ATR方法[J]. 现代雷达, 2004, 26 (5): 20- 23. doi: 10.3969/j.issn.1004-7859.2004.05.006
	GAO Q , LIU J X . HRRP-ATR based on SVM classifier[J]. Modern Radar, 2004, 26 (5): 20- 23. doi: 10.3969/j.issn.1004-7859.2004.05.006
4	余责永, 刘汪渡, 谢永亮. 基于SVM的雷达目标一维距离像识别[J]. 舰船电子工程, 2009, 29 (9): 102- 104.
	YU Z Y , LIU W D , XIE Y L . Radar target recognition using range profiles based on SVM[J]. Ship Electronic Engineering, 2009, 29 (9): 102- 104.
5	DU L , LIU H W , BAO Z , et al. Radar automatic target reco-gnition using complex high-resolution range profiles[J]. IET Radar Sonar Navigation, 2007, 1 (1): 18- 26. doi: 10.1049/iet-rsn:20050119
6	张小英, 王宝发, 刘铁军. 基于PCA-LVQ的雷达目标一维距离像识别[J]. 系统工程与电子技术, 2005, 27 (8): 1373- 1375. doi: 10.3321/j.issn:1001-506X.2005.08.013
	ZHANG X Y , WANG B F , LIU T J . 1-D range profile recognition of radar targets using PCA-LVQ[J]. Systems Engineering and Electronics, 2005, 27 (8): 1373- 1375. doi: 10.3321/j.issn:1001-506X.2005.08.013
7	李波, 任红梅, 肖志河. 人工智能在雷达应用中的限制和发展前景[J]. 军事文摘, 2019, 435 (3): 42- 44.
	LI B , REN H M , XIAO Z H . Limitation and development prospect of artificial intelligence in radar application[J]. Military Digest, 2019, 435 (3): 42- 44.
8	徐高贵, 殷红成, 袁莉, 等. 一种基于卷积神经网络的HRRP序列空间目标识别方法[J]. 中国传媒大学学报(自然与科学版), 2019, 26 (3): 40- 44.
	XU G G , YIN H C , YUAN L , et al. Spatial target recognition method of HRRP sequence based on convolutional neural network[J]. Journal of Communication University of China (Science and Technology), 2019, 26 (3): 40- 44.
9	王容川, 庄志洪, 王宏波, 等. 基于卷积神经网络的雷达目标HRRP分类识别方法[J]. 现代雷达, 2019, 41 (5): 33- 38.
	WANG R C , ZHUANG Z H , WANG H B , et al. HRRP classification and recognition method of radar target based on convolutional neural network[J]. Modern Radar, 2019, 41 (5): 33- 38.
10	陆金文, 殷红成, 盛晶, 等. 基于改进的一维卷积神经网络的高分辨距离像识别方法[J]. 电光与控制, 2020, 27 (8): 19- 22.
	LU J W , YIN H C , SHENG J , et al. High-resolution range profile recognition based on improved one-dimensional convolutional neural network[J]. Electronics Optics & Control, 2020, 27 (8): 19- 22.
11	JARMO L, VISA K. Deep learning for HRRP-based target recognition in multistatic radar systems[C]//Proc. of the IEEE Radar Conference, 2016.
12	冯博. 雷达高分辨距离像特征提取与识别方法研究[D]. 西安: 西安电子科技大学, 2015.
	FENG B. Feature extraction and recognition methods for radar high range resolution profiles[D]. Xi'an: Xidian University, 2015.
13	冯博, 陈渤, 王鹏辉, 等. 基于稳健深层网络的雷达高分辨距离像目标特征提取算法[J]. 电子与信息学报, 2014, 36 (12): 2949- 2955.
	FENG B , CHEN B , WANG P H , et al. Feature extraction method for radar high resolution range profile targets based on robust deep networks[J]. Journal of Electronics & Information Technology, 2014, 36 (12): 2949- 2955.
14	翟颖, 陈渤. 基于稳健变分自编码器模型的雷达高分辨距离像目标识别算法[J]. 电子学报, 2020, 48 (6): 1149- 1150.
	ZHAI Y , CHEN B . Robust variational auto-encoder for radar HRRP target recognition[J]. Acta Electronica Sinica, 2020, 48 (6): 1149- 1150.
15	张增, 任汉付. 基于SAE-HMM模型的雷达目标高分辨率距离像识别方法[P]. 中国: CN 109871907 A, 20190611.
	ZHANG Z, REN H F. Radar target high resolution range profile recognition method based on SAE-HMM model[P]. China: CN 109871907 A, 20190611.
16	CHENG X Y , PENG X Y , LI J B , HUO C Y . Sparse autoencoder based manifold analyzer model of multi-angle target feature[J]. IEEE Access, 2020, 8, 153250- 153263.
17	BYRD R H , NOCEDAL J , SCHNABEL R B . Representations of quasi-Newton matrices and their use in limited memory me-thods[J]. Mathematical Programming, 1994, 63 (1/3): 129- 156.
18	肖慧. 物理光学电磁建模方法的研究与实现[D]. 长沙: 国防科学技术大学, 2003.
	XIAO H. Research on and implementation of electromagnetic modeling method based on physical optics[D]. Changsha: National University of Defense Technology, 2003.
19	崔索民, 吴振森, 方大纲. 一种计算多面体目标RCS的等效边缘电磁流公式[J]. 微波学报, 1997, 13 (1): 20- 25.
	CUI S M , WU Z S , FANG D G . A novel formula of equivalent edge current for calculating RCS of polyhedral structure[J]. Journal of Microwaves, 1997, 13 (1): 20- 25.
20	殷红成, 朱国庆, 董纯柱, 等. 基于自适应射线管分裂的多次反射计算方法[J]. 系统工程与电子技术, 2013, 35 (4): 700- 706.
	YIN H C , ZHU G Q , DONG C Z , et al. Efficient multi-reflection computational method based on adaptive ray tube splitting[J]. Systems Engineering and Electronics, 2013, (4): 700- 706.
21	QUINQUIS A, DEMETER S, RADOIE. Enhancing the resolution of radar range profiles using a class of subspace eigenanalysis based techniques[C]//Proc. of the Digital Signal Processing, 2001: 288-303.

标准体名称	标准体尺寸	坐标中心	建模方法
三面角	直角边长 600 mm、壁厚5 mm	直角中心位于(5,0,0)	PO+EEC+ SBR(3次反射)
二面角	内板尺寸 1 m×0.5 m、壁厚5 mm	直角中心位于(0,0,0)	PO+EEC+ SBR(2次反射)
立方体	边长1 m	直角中心位于(0,0,0)	PO+EEC

降维维数	冗余度/%
256	80.4
64	65.6
16	56.2
15	53.3
14	50.0
13	46.1
12	41.6
11	45.4
10	40.0
9	33.3
8	37.5
7	28.6
6	33.3
5	40.0

[1]	丁达理, 魏政磊, 唐上钦, 黄长强. 自适应预测权重的空战鲁棒机动决策方法[J]. 系统工程与电子技术, 2020, 42(10): 2275-2284.
[2]	孟克子, 周荻. 多约束条件下的最优中制导律设计[J]. 系统工程与电子技术, 2016, 38(1): 116-122.
[3]	肖冰松,方洋旺,许蕴山,王磊,吴宗一. 基于模糊集和粗糙集的武器系统作战能力评估[J]. Journal of Systems Engineering and Electronics, 2010, 32(6): 1263-1265.

HRRP稀疏自编码器深层特征与散射中心特征的关联性研究

Correlation research between deep features of HRRP sparse auto-encoder and scattering center features

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 22

参考文献 21

相关文章 3

编辑推荐

Metrics

本文评价