基于AlexNet迁移学习的磁异常信号检测方法

李启飞; 吴 芳; 韩蕾蕾; 范赵鹏; 李沛宗

doi:10.11993/j.issn.2096-3920.2020.02.007

基于AlexNet迁移学习的磁异常信号检测方法

doi: 10.11993/j.issn.2096-3920.2020.02.007

1.
海军航空大学战勤学院, 山东烟台, 264001
2.
中国人民解放军91550部队, 辽宁大连, 116000
3.
中国人民解放军91001部队, 北京, 100080

基金项目: 国家自然科学基金(61971424)

详细信息

作者简介:
李启飞(1994-), 男, 在读硕士, 主要从事磁信号处理相关研究.

中图分类号: TJ67; TM937
计量
- 文章访问数: 387
- HTML全文浏览量: 1
- PDF下载量: 191
- 被引次数: 0
出版历程
- 收稿日期: 2018-12-06
- 修回日期: 2019-01-04
- 刊出日期: 2020-04-30

Detection Method of Magnetic Anomaly Signals Based on AlexNet Transfer Learning

1.
Air Combat Service College, Naval Aviation University, Yantai 264001, China
2.
91550th Unit, the People’s Liberation Army of China, Dalian 116000, China
3.
91001th Unit, the People’s Liberation Army of China, Beijing 100080, China

摘要

摘要: 航空磁探反潜是通过磁探仪检测磁异常信号, 进行航空反潜的重要手段。针对水下目标的磁异常信号在低信噪比条件下难以检测的问题, 提出使用基于AlexNet迁移学习的磁异常信号检测方法。该方法基于卷积神经网络原理, 通过对水下目标不同态势、航空磁探平台不同飞行路线和速度情况进行仿真, 得到大量信号序列, 对其加高斯白噪声, 从而模拟磁探平台测得信号。随后对其进行短时傅里叶变换得到时频图, 并利用深度卷积神经网络模型AlexNet对时频图特征进行迁移学习训练, 最后利用测试集数据对训练后的AlexNet网络进行测试, 实现对低信噪比条件下水下目标磁异常信号的检测。仿真结果表明, 在信噪幅度比为-8 dB、虚警率为3%情况下, 对磁异常信号的检测概率达到93%。
- 航空反潜 /
- 磁异常信号 /
- 检测 /
- 时频分析 /
- AlexNet /
- 迁移学习
Abstract: Aiming at the problem in aviation antisubmarine that the magnetic anomaly signals of underwater targets are difficult to detect under low signal-to-noise ratio(SNR), a detection method of magnetic anomaly signals based on AlexNet transfer learning is proposed. Based on the principle of convolution neural network, a large number of signal sequences are obtained by simulating different situations of underwater targets and different flight routes and speeds of the aeromagnetic platform with Gaussian white noise, so as to simulate the signals measured by the magnetic detection platform. Then, the short-time Fourier transform(STFT) is applied to obtain the time-frequency diagram, and the time-frequency features are transferred and trained by using the AlexNet. Finally, the trained AlexNet deep convolutional neural network is tested by the test set. The results show that the AlexNet model has good generalization ability in the field of magnetic anomaly signal detection. When the SNR is -8 dB and the false alarm rate is 3%, the detection probability of magnetic anomaly signal reaches 93%.
- aviation antisubmarine /
- magnetic anomaly signal /
- detection /
- time-frequency analysis /
- AlexNet /
- transfer learning

HTML全文

参考文献(14)

[1]	林君, 刁庶, 张洋, 等. 地球物理矢量场磁测技术的研究进展[J]. 科学通报, 2017, 62(23): 2606-2618. Lin Jun, Diao Shu, Zhang Yang, et al. Research Progress of Geophysical Vector Magnetic Field Survey Technology[J]. Chin Sci Bull, 2017, 62(23): 2606-2618.
[2]	周家新, 陈建勇, 单志超, 等. 航空磁探中潜艇目标的联合估计检测方法研究[J]. 兵工学报, 2018, 39(5): 833- 840. Zhou Jia-xin, Chen Jian-yong, Shan Zhi-chao, et al. Research on Joint Estimation and Detection of Submarine Target in Airborne Magnetic Anomaly Detection[J]. Acta Armamentarii, 2018, 39(5): 833-840.
[3]	Ginzburg B, Frumkis L, Kaplan Z, et al. Processing of Magnetic Scalar Gradiometer Signals Using Orthonormalized Functions[J]. Sensors, 2002, 102(1): 67-75.
[4]	Sheinker A, Frumkis L, Ginzburg B, et al. Magnetic Anomaly Detection Using a Three-Axis Magnetometer[J]. IEEE Transactions on Magnetics, 2009, 45(1): 160-167.
[5]	张坚, 林春生, 黄凡. OBF分解与BP网络在船舶磁场信号检测中的应用[J]. 船电技术, 2011, 31(7): 13-16. Zhang Jian, Lin Chun-sheng, Huang Fan. Application of OBF Decomposition and BP Neural Network to Magnetic Signal Detection of a Ship[J]. Marine Electric ＆ Electronic Technology, 2011, 31(7): 13-16.
[6]	张坚, 林春生, 邓鹏, 等. 基于小波域OBF分解的磁异常信号检测算法[J]. 弹箭与制导学报, 2011, 31(6): 187-189. Zhang Jian, Lin Chun-sheng, Deng Peng, et al. Detection of Magnetic Anomaly Signal Based on OBF Decomposition in Wavelet Domain[J]. Journal of Projectiles, Rockets, Mi- ssiles and Guidance, 2011, 31(6): 187-189.
[7]	王珺琳, 陈博. 基于运动目标的OBF方法检测性能分析[J]. 中国电子科学研究院学报, 2017, 12(1): 47-51. Wang Jun-lin, Chen Bo. The Analysis of OBF Detection Performance Based on Moving Object[J]. Journal of China Academy of Electronics and Information Technology, 2017, 12(1): 47-51.
[8]	Wan C, Pan M, Zhang Q, et al. Magnetic Anomaly Detection Based on Stochastic Resonance[J]. Sensors Actuators A Physical, 2018, 278: 11-17.
[9]	Sheinker A, Moldwin M B. Magnetic Anomaly Detection(MAD) of Ferromagnetic Pipelines Using Principal Component Analysis(PCA)[J]. Measurement Science and Technology, 2016, 27(4): 045104.
[10]	Sheinker A, Salomonski N, Ginzburg B, et al. Magnetic Anomaly Detection Using Entropy Filter[J]. Measurement Science, 2008, 19(4): 045205.
[11]	Sheinker A, Ginzburg B, Salomonski N, et al. Magnetic Anomaly Detection Using High-Order Crossing Method [J]. Transactions on Geoscience, 2012, 50(4): 1095-1103.
[12]	Krizhevsky A, Sutskever I, Hinton G E. ImageNet Classi-fication with Deep Convolutional Neural Net-works[C]//NIPS’12: Proceedings of the 27th International Conference on Neural Information Processing Systems. USA: Massachusetts Institute of Technology Press, 2012, 1: 1097-1105.
[13]	Yosinski J, Clune J, Bengio Y, et al. How Transferable are Features in Deep Neural Networks[C]//NIPS’14: Proceedings of the 27th International Conference on Neural Information Processing Systems. USA: Massachusetts Institute of Technology Press, 2014, 2: 3320-3328.
[14]	Cun Y L, Boser B, Denker J S, et al. Handwritten Digit Recognition with a Back-Propagation Network[C]//Advances in Neural Information Processing Systems. USA: Massachusetts Institute of Technology Press, 1990, 2: 396-404.