基于栈式降噪稀疏自动编码器的雷达目标识别方法

赵飞翔; 刘永祥; 霍凯

doi:10.12000/JR16151

基于栈式降噪稀疏自动编码器的雷达目标识别方法

DOI: 10.12000/JR16151

国防科学技术大学电子科学与工程学院长沙 410073

基金项目: 国家自然科学基金优秀青年基金(61422114)，湖南省杰出青年科学基金(2015JJ1003)

详细信息

作者简介:
赵飞翔(1989–)，男，河南洛阳人，国防科学技术大学电子科学与工程学院在读博士生，研究方向为雷达目标识别。E-mail: zfxkj123@sina.cn

刘永祥(1976–)，男，河北唐山人，博士，国防科学技术大学电子科学与工程学院教授，博士生导师，主要研究方向为目标微动特性分析与识别。E-mail: lyx_bible@sina.com

霍凯：霍凯(1983–)，男，湖北黄冈人，博士，国防科学技术大学电子科学与工程学院讲师，主要研究方向为雷达信号处理与目标识别。E-mail: huokai2001@163.com

通讯作者:
刘永祥 lyx_bible@sina.com

中图分类号: TN957.51
计量
- 文章访问数: 3260
- HTML全文浏览量: 865
- PDF下载量: 1331
- 被引次数: 0
出版历程
- 收稿日期: 2016-12-22
- 修回日期: 2017-01-10
- 网络出版日期: 2017-04-28

Radar Target Recognition Based on Stacked Denoising Sparse Autoencoder

College of Electronic Science and Engineering, National University of Defense Technology, Changsha 410073, China

Funds: The National Natural Science Foundation of China (61422114), The Natural Science Fund for Distinguished Young Scholars of Hunan Province (2015JJ1003)

摘要

摘要: 雷达目标识别中特征提取是关键步骤，所提取特征的好坏决定着识别效果的优劣，但传统特征提取方法很难发掘目标数据深层次本质特征。深度学习理论中的自动编码器模型能够用数据去学习特征，获得数据不同层次的特征表达。同时为消除噪声影响，该文提出一种基于栈式降噪稀疏自动编码器的雷达目标识别方法，通过设置不同隐藏层数和迭代次数，从雷达数据中直接高效地提取识别所需的各层次特征。暗室仿真数据实验结果验证了该方法较K近邻分类方法及传统栈式自编码器有更好的识别效果。
- 目标识别 /
- 深度学习 /
- 栈式降噪稀疏自动编码器
Abstract: Feature extraction is a key step in radar target recognition. The quality of the extracted features determines the performance of target recognition. However, obtaining the deep nature of the data is difficult using the traditional method. The autoencoder can learn features by making use of data and can obtain feature expressions at different levels of data. To eliminate the influence of noise, the method of radar target recognition based on stacked denoising sparse autoencoder is proposed in this paper. This method can extract features directly and efficiently by setting different hidden layers and numbers of iterations. Experimental results show that the proposed method is superior to the K-nearest neighbor method and the traditional stacked autoencoder.
- Target recognition /
- Deep learning /
- Stacked denoising sparse autoencoder

HTML全文

图 1 自动编码器结构

Figure 1. The structure of AE

下载: 全尺寸图片幻灯片

图 2 降噪稀疏自动编码器结构

Figure 2. The structure of DSAE

下载: 全尺寸图片幻灯片

图 3 使用sDSAE进行识别分类的完整过程图

Figure 3. The process of using sDSAE to identify targets

下载: 全尺寸图片幻灯片

图 4 目标结构示意图

Figure 4. Schematic diagram of target structure

下载: 全尺寸图片幻灯片

图 5 3类目标RCS 3维效果图

Figure 5. Three dimensional effect chart of RCS of the three-class targets

下载: 全尺寸图片幻灯片

图 6 sAE与sDSAE识别结果对比图

Figure 6. Comparison of recognition results based on sAE and sDSAE

下载: 全尺寸图片幻灯片

图 7 不同代价函数对识别效果的影响

Figure 7. The effect of cost function on recognition

下载: 全尺寸图片幻灯片

图 8 不同信噪比下sAE与sDSAE识别结果对比图

Figure 8. Comparison of recognition results based on sAE and sDSAE under different SNR conditions

下载: 全尺寸图片幻灯片

表 1 选用不同特征K近邻分类方法结果

Table 1. Recognition results based on K-nearest neighbor method

特征	识别准确率
均值	0.9503
方差	0.4811
梅林变换均值	0.8519
梅林变换方差	0.7427
均值、梅林变换均值	0.9011
均值、方差、梅林变换方差	0.7247
均值、梅林变换均值、梅林变换方差	0.8483
均值、方差、梅林变换均值、梅林变换方差	0.7565

下载: 导出CSV

表 2 采用sDSAE识别结果

Table 2. Recognition results based on sDSAE

隐藏层	迭代10次	迭代30次	迭代50次
1	0.6254	0.6992	0.7185
2	0.7998	0.8713	0.9047
3	0.8763	0.9247	0.9511
4	0.9124	0.9458	0.9679

下载: 导出CSV

表 3 不同信噪比下采用sDSAE识别结果

Table 3. Recognition results based on sDSAE under different SNR conditions

信噪比(dB)	迭代次数	隐藏层
信噪比(dB)	迭代次数	1	2	3	4
20	迭代10次	0.6123	0.7987	0.8925	0.9006
	迭代30次	0.6395	0.8604	0.9105	0.9246
	迭代50次	0.6654	0.8816	0.9233	0.9412
10	迭代10次	0.5757	0.6788	0.7491	0.8499
	迭代30次	0.5967	0.7212	0.8027	0.8756
	迭代50次	0.6019	0.7964	0.8259	0.9011
5	迭代10次	0.5011	0.6679	0.7368	0.8004
	迭代30次	0.5548	0.7012	0.7449	0.8368
	迭代50次	0.5952	0.7229	0.7866	0.8514

下载: 导出CSV

参考文献(18)

[1]	吴剑旗, 田西兰. 一种基于半监督学习的窄带雷达目标识别系统[J]. 中国电子科学研究院学报, 2015, 10(1): 49–53. http://www.cnki.com.cn/Article/CJFDTOTAL-KJPL201501008.htm Wu Jian-qi and Tian Xi-lan. A narrow-band radar target recognition system based on semi-supervised learning[J].Journal of CAEIT, 2015, 10(1): 49–53. http://www.cnki.com.cn/Article/CJFDTOTAL-KJPL201501008.htm
[2]	张平定, 孙佳佳, 童创明, 等. 弹道中段目标雷达综合识别研究[J]. 微波学报, 2015, 31(2): 20–23. http://cdmd.cnki.com.cn/Article/CDMD-90002-1012020687.htm Zhang Ping-ding, Sun Jia-jia, Tong Chuang-ming, et al. Integrated target recognition of ballistic midcourse target[J].Journal of Microwaves, 2015, 31(2): 20–23. http://cdmd.cnki.com.cn/Article/CDMD-90002-1012020687.htm
[3]	曹伟, 周智敏, 周辉, 等. 基于多维特征及BP网络的高分辨雷达目标识别[J]. 计算机工程与应用, 2013, 49(8): 213–216. http://www.cnki.com.cn/Article/CJFDTOTAL-JSGG201308052.htm Cao Wei, Zhou Zhi-min, Zhou Hui, et al. High resolution radar target recognition based on multi-dimensional feature vector and BP network[J].Computer Engineering and Applications, 2013, 49(8): 213–216. http://www.cnki.com.cn/Article/CJFDTOTAL-JSGG201308052.htm
[4]	Hinton G E and Salakhutdinov R R. Reducing the dimensionality of data with neural networks[J].Science, 2006, 313(5786): 504–507. doi: 10.1126/science.1127647
[5]	Lecun Y, Bengio Y, and Hinton G. Deep learning[J].Nature, 2015, 521(7553): 436–444. doi: 10.1038/nature14539
[6]	尹宝才, 王文通, 王立春. 深度学习研究综述[J]. 北京工业大学学报, 2015, 41(1): 48–57. http://youxian.cnki.com.cn/yxdetail.aspx?filename=DSWZ201706161&dbname=CJFDPREN Yin Bao-cai, Wang Wen-tong, and Wang Li-chun. Review of deep learning[J].Journal of Beijing University of Technology, 2015, 41(1): 48–57. http://youxian.cnki.com.cn/yxdetail.aspx?filename=DSWZ201706161&dbname=CJFDPREN
[7]	丁军, 刘宏伟, 陈渤, 等. 相似性约束的深度置信网络在SAR图像目标识别的应用[J]. 电子与信息学报, 2016, 38(1): 97–103. http://www.cnki.com.cn/Article/CJFDTOTAL-DZYX201601013.htm Ding Jun, Liu Hong-wei, Chen Bo, et al. Similarity constrained deep belief networks with application to SAR image target recognition[J].Journal of Electronics &Information Technology, 2016, 38(1): 97–103. http://www.cnki.com.cn/Article/CJFDTOTAL-DZYX201601013.htm
[8]	田壮壮, 占荣辉, 胡杰民, 等. 基于卷积神经网络的SAR图像目标识别研究[J]. 雷达学报, 2016, 5(3): 320–325. http://www.cnki.com.cn/Article/CJFDTOTAL-LDAX201603012.htm Tian Zhuang-zhuang, Zhan Rong-hui, Hu Jie-min, et al. SAR ATR based on convolutional neural network[J].Journal of Radars, 2016, 5(3): 320–325. http://www.cnki.com.cn/Article/CJFDTOTAL-LDAX201603012.htm
[9]	Jiang Xiao-juan, Zhang Ying-hua, Zhang Wen-sheng, et al. A novel sparse auto-encoder for deep unsupervised learning[C]. 2013 Sixth International Conference on Advanced Computational Intelligence, Hangzhou, 2013: 256–261.
[10]	Feng Bo, Chen Bo, and Liu Hong-wei. Radar HRRP target recognition with deep networks[J].Pattern Recognition, 2017, 61: 379–393.
[11]	张成刚, 姜静清. 一种稀疏降噪自编码神经网络研究[J]. 内蒙古民族大学学报(自然科学版), 2016, 31(1): 21–25. http://www.cnki.com.cn/Article/CJFDTOTAL-NMMS201601007.htm Zhang Cheng-gang and Jiang Jing-qing. Study on sparse De-noising Auto-Encoder neural network[J].Journal of Inner Mongolia University for Nationalities, 2016, 31(1): 21–25. http://www.cnki.com.cn/Article/CJFDTOTAL-NMMS201601007.htm
[12]	Meng Ling-heng, Ding Shi-fei, and Xue Yu. Research on denoising sparse autoencoder[J].International Journal of Machine Learning and Cybernetics, 2016. DOI: 10.1007/s13042-016-0550-y.
[13]	Kumar V, Nandi G C, and Kala R. Static hand gesture recognition using stacked denoising sparse autoencoders[C]. 2014 Seventh International Conference on Contemporary Computing, Noida, 2014: 99–104.
[14]	Sankaran A, Pandey P, Vatsa M, et al. On latent fingerprint minutiae extraction using stacked denoising sparse AutoEncoders[C]. IEEE International Joint Conference on Biometrics, Clearwater, FL, 2014: 1–7.
[15]	Jose Dolz, Nacim Betrouni, Mathilde Quidet, et al. Stacking denoising auto-encoders in a deep network to segment the brainstem on MRI in brain cancer patients: A clinical study[J].Computerized Medical Imaging and Graphics, 2016, 52: 8–18. doi: 10.1016/j.compmedimag.2016.03.003
[16]	Michael A Nielsen. Neural Networks and Deep Learning[M]. Determination Press, 2015.
[17]	Sun Wen-jun, Shao Si-yu, Zhao Rui, et al. A sparse auto-encoder-based deep neural network approach for induction motor faults classification[J].Measurement, 2016, 89: 171–178. doi: 10.1016/j.measurement.2016.04.007
[18]	Xing Chen, Ma Li, and Yang Xiao-quan. Stacked denoise autoencoder based feature extraction and classification for hyperspectral images[J].Journal of Sensors, 2016. DOI: 10.1155/2016/3632943.