轻量化CNN的小样本目标电场线谱检测算法

刘琪; 郑伟

doi:10.7643/ issn.1672-9242.2025.09.008

PDF(5223 KB)

装备环境工程 ›› 2025, Vol. 22 ›› Issue (9) : 68-77. DOI: 10.7643/ issn.1672-9242.2025.09.008

专题——舰船装备可靠性

轻量化CNN的小样本目标电场线谱检测算法

刘琪¹, 郑伟²

作者信息 +

Lightweight Convolutional Neural Network Based Algorithm for Target Electric Field Line Spectrum Detection in Small Sample Size Scenarios

LIU Qi¹, ZHENG Wei²

Author information +

文章历史 +

摘要

目的实现复杂海洋噪声干扰及小样本条件下目标轴频电场线谱的有效检测。方法采用物理引导的数据增强机制,设计频移缩放、能量调制与噪声扰动策略生成仿真数据。构建轻量化多尺度CNN架构,引入深度可分离卷积与频移注意力模块,并运用跨域迁移优化策略实现知识迁移。结果在含10组实测与200组仿真数据的混合训练集上验证算法性能。在-5 dB信噪比下,可完整检测10 s时长目标信号的主要频率线谱。与5种传统方法对比,所提算法的线谱检测均方误差仅0.001 5,平均绝对误差仅0.032 Hz,0.1 Hz准确率达96.1%,在检测精度和鲁棒性上均更优,综合效果更好。结论本文深度融合物理建模与深度学习,解决了复杂海洋噪声下小样本轴频电场信号的有效检测问题,为水下目标探测提供了一种高鲁棒、低数据依赖的解决方案,同时为后续多物理场联合感知和在线增量学习机制的探索奠定了理论基础。

Abstract

The work aims to achieve effective detection of target axial frequency electric field line spectra under complex ocean noise interference and small sample conditions. A physically guided data augmentation mechanism was employed to generate simulation data by frequency-shift scaling, energy modulation, and noise perturbation strategies. A lightweight multi-scale CNN architecture was constructed, incorporating depth-wise separable convolution and frequency-shift attention modules. Furthermore, a cross-domain transfer optimization strategy was utilized to facilitate knowledge transfer. The algorithm's performance was verified on a hybrid training set comprising 10 groups of real-world data and 200 groups of simulated data. At a signal-to-noise ratio (SNR) of -5 dB, the complete detection of the line spectra of target signal's main frequencies within a 10 s duration could be achieved. Compared with five traditional methods, the proposed algorithm demonstrated a line spectrum detection mean squared error (MSE) of just 0.001 5, an average absolute error of only 0.032 Hz, and a 0.1 Hz accuracy of 96.1%, showing superior detection accuracy and robustness, as well as a better overall performance. This paper presents a deep integration of physical modeling and deep learning, offering an effective solution for detecting small-sample axial frequency electric field signals in the presence of complex ocean noise. It provides a highly robust and low-data-dependent approach for underwater target detection, while also laying a theoretical foundation for future exploration of multi-physical field joint perception and online incremental learning mechanisms.

导出引用

刘琪, 郑伟. 轻量化CNN的小样本目标电场线谱检测算法[J]. 装备环境工程. 2025, 22(9): 68-77 https://doi.org/10.7643/ issn.1672-9242.2025.09.008

LIU Qi, ZHENG Wei. Lightweight Convolutional Neural Network Based Algorithm for Target Electric Field Line Spectrum Detection in Small Sample Size Scenarios[J]. Equipment Environmental Engineering. 2025, 22(9): 68-77 https://doi.org/10.7643/ issn.1672-9242.2025.09.008

中图分类号： TP183

参考文献

[1] 喻鹏, 程锦房, 张伽伟, 等. 基于Rao检测器的舰船轴频电场滑动门限检测方法[J]. 兵工学报, 2021, 42(4): 827-834.
YU P, CHENG J F, ZHANG J W, et al.Ship Shaft-Rate Electric Field Sliding Threshold Detection Method Based on Rao Detector[J]. Acta Armamentarii, 2021, 42(4): 827-834.
[2] 程锦房, 喻鹏, 张伽伟, 等. 水下电场探测定位技术应用研究现状[J]. 海军工程大学学报, 2022, 34(4): 68-74.
CHENG J F, YU P, ZHANG J W, et al.Application and Development of Underwater Electric Field Detection and Location Technology[J]. Journal of Naval University of Engineering, 2022, 34(4): 68-74.
[3] 程锦房, 张伽伟, 姜润翔, 等. 水下电磁探测技术的发展现状[J]. 数字海洋与水下攻防, 2019, 2(4): 45-49.
CHENG J F, ZHANG J W, JIANG R X, et al.Development Status of Underwater Electromagnetic Detection Technology[J]. Digital Ocean & Underwater Warfare, 2019, 2(4): 45-49.
[4] 赵文春, 姜润翔, 喻鹏, 等. 基于轴频电场线谱特征的目标检测及识别[J]. 兵工学报, 2020, 41(6): 1165-1171.
ZHAO W C, JIANG R X, YU P, et al.Detection and Identification of Ship Shaft-Rate Electric Field Based on Line-Spectrum Characteristics[J]. Acta Armamentarii, 2020, 41(6): 1165-1171.
[5] 金志浩, 张旭, 张义民, 等. Welch功率谱与卷积神经网络结合的滚动轴承故障诊断[J]. 机械设计与制造, 2024(2): 271-275.
JIN Z H, ZHANG X, ZHANG Y M, et al.Fault Diagnosis of Rolling Bearing Based on Welch Power Spectrum with Convolution Neural Network[J]. Machinery Design & Manufacture, 2024(2): 271-275.
[6] 张旭, 徐永海, 秦本双, 等. 基于正交匹配追踪算法的谐波源定位方法[J]. 电测与仪表, 2021, 58(4): 44-51.
ZHANG X, XU Y H, QIN B S, et al.Harmonic Source Localization Method Based on Orthogonal Matching Pursuit Algorithm[J]. Electrical Measurement & Instrumentation, 2021, 58(4): 44-51.
[7] 李彦楠, 史明佳, 王伟, 等. 弱目标自动跟踪中的线谱检测[J]. 太赫兹科学与电子信息学报, 2024, 22(8): 901-907.
LI Y N, SHI M J, WANG W, et al.Weak Target Tracking Algorithm Based on Line Spectrum Detection[J]. Journal of Terahertz Science and Electronic Information Technology, 2024, 22(8): 901-907.
[8] 陈云杰, 柏溢. 基于卷积神经网络的移动通信网络入侵检测方法研究[J]. 无线互联科技, 2024, 21(17): 122-124.
CHEN Y J, BAI Y.Study on the Intrusion Detection Method of Mobile Communication Network Based on Convolutional Neural Network[J]. Wireless Internet Science and Technology, 2024, 21(17): 122-124.
[9] 马世拓, 班一杰, 戴陈至力. 卷积神经网络综述[J]. 现代信息科技, 2021, 5(2): 11-15.
MA S T, BAN Y J, DAI C.Survey of Convolutional Neural Network[J]. Modern Information Technology, 2021, 5(2): 11-15.
[10] 王亚辉, 余隋怀, 陈登凯, 等. 基于深度学习的人工智能设计决策模型[J]. 计算机集成制造系统, 2019, 25(10): 2467-2475.
WANG Y H, YU S H, CHEN D K, et al.Artificial Intelligence Design Decision Making Model Based on Deep Learning[J]. Computer Integrated Manufacturing Systems, 2019, 25(10): 2467-2475.
[11] 居云龙, 周怡君, 罗晨, 等. 基于改进DenseNet模型的船舶图像分类[J]. 船舶标准化工程师, 2024, 57(5): 80-86.
JU Y L, ZHOU Y J, LUO C, et al.Ship Image Classification Based on Improved Densenet Model[J]. Ship Standardization Engineer, 2024, 57(5): 80-86.
[12] 周云, 赵瑜, 郝官旺, 等. 基于应变信号时频分析与CNN网络的车辆荷载识别方法[J]. 湖南大学学报(自然科学版), 2022, 49(1): 21-32.
ZHOU Y, ZHAO Y, HAO G W, et al.Vehicle Load Identification Method Based on Time Frequency Analysis of Strain Signal and Convolutional Neural Network[J]. Journal of Hunan University (Natural Sciences), 2022, 49(1): 21-32.
[13] 袁建虎, 韩涛, 唐建, 等. 基于小波时频图和CNN的滚动轴承智能故障诊断方法[J]. 机械设计与研究, 2017, 33(2): 93-97.
YUAN J H, HAN T, TANG J, et al.An Approach to Intelligent Fault Diagnosis of Rolling Bearing Using Wavelet Time-Frequency Representations and CNN[J]. Machine Design & Research, 2017, 33(2): 93-97.
[14] LEE D, JEONG J.Few-Shot Learning-Based Light- Weight WDCNN Model for Bearing Fault Diagnosis in Siamese Network[J]. Sensors (Basel), 2023, 23(14): 6587.
[15] LIU Z, LIN Y T, CAO Y, et al.Swin Transformer: Hierarchical Vision Transformer Using Shifted Windows[C]// 2021 IEEE/CVF International Conference on Computer Vision (ICCV). Montreal, QC, Canada. IEEE, 2021: 9992-10002.
[16] FE-FEI L, Fergus, Perona.A Bayesian Approach to Unsupervised One-Shot Learning of Object Categories[C]// Proceedings Ninth IEEE International Conference on Computer Vision. Nice, France. IEEE, 2003: 1134-1141.
[17] 王凯, 任劼, 章为川. 基于Swin Transformer的图神经网络小样本图像分类算法[J]. 激光与光电子学进展, 2024, 61(12): 381-389.
WANG K, REN J, ZHANG W C.Few-Shot Image Classification Algorithm of Graph Neural Network Based on Swin Transformer[J]. Laser & Optoelectronics Progress, 2024, 61(12): 381-389.
[18] WANG T, CHEN C Z, DONG X J, et al.A Novel Method of Production Line Bearing Fault Diagnosis Based on 2D Image and Cross-Domain Few-Shot Learning[J]. Applied Sciences, 2023, 13(3): 1809.
[19] 李旭冬, 叶茂, 李涛. 基于卷积神经网络的目标检测研究综述[J]. 计算机应用研究, 2017, 34(10): 2881-2886.
LI X D, YE M, LI T.Review of Object Detection Based on Convolutional Neural Networks[J]. Application Research of Computers, 2017, 34(10): 2881-2886.
[20] 胡鹏, 肖爱明, 龚沈光. 基于小波尺度相关的船舶轴频电场信号检测算法[J]. 探测与控制学报, 2013, 35(2): 37-40.
HU P, XIAO A M, GONG S G.Detection Algorithm for Ship Shaft-Rate Electric Field Signal Based on Wavelet Scale Correlation[J]. Journal of Detection & Control, 2013, 35(2): 37-40.
[21] 刘俊利, 缪炳荣, 张盈, 等. 基于Morlet小波与CART决策树的滚动轴承故障诊断方法[J]. 机械强度, 2024, 46(1): 1-8.
LIU J L, MIAO B R, ZHANG Y, et al.Rolling Bearing Fault Diagnosis Method Based on Morlet Wavelet and Cart Decision Tree[J]. Journal of Mechanical Strength, 2024, 46(1): 1-8.
[22] 苑传林. 基于FFT信号处理器的舰船雷达目标检测[J]. 舰船科学技术, 2018, 40(8): 76-78.
YUAN C L.Ship Radar Target Detection Based on FFT Signal Processor[J]. Ship Science and Technology, 2018, 40(8): 76-78.
[23] 左金辉, 贾豫东, 张晓青, 等. 相干多普勒测风激光雷达的频率估计算法[J]. 激光与红外, 2021, 51(5): 554-558.
ZUO J H, JIA Y D, ZHANG X Q, et al.Frequency Estimation Algorithm of Coherent Doppler Wind Lidar[J]. Laser & Infrared, 2021, 51(5): 554-558.
[24] 马亚男, 戴尔晗, 陈诚. 一种利用自相关法和三点法检测信号频率的方法[J]. 微型机与应用, 2016, 35(14): 74-76.
MA Y N, DAI E H, CHEN C.Method for Detecting Signal Frequency by Using Self-Correlation Method and Three Point Method[J]. Microcomputer & Its Applications, 2016, 35(14): 74-76.
[25] 朱琴跃, 于逸尘, 占岩文, 等. 基于短时傅里叶变换和深度网络的模块化多电平换流器子模块IGBT开路故障诊断[J]. 电工技术学报, 2024, 39(12): 3840-3854.
ZHU Q Y, YU Y C, ZHAN Y W, et al.IGBT Open-Circuit Fault Diagnosis of Modular Multilevel Converter Sub-Module Based on Short-Time Fourier Transform and Deep Networks[J]. Transactions of China Electrotechnical Society, 2024, 39(12): 3840-3854.
[26] 单广超, 赵汉波. 舰船辐射噪声线谱检测与分析[J]. 舰船电子工程, 2014, 34(10): 119-122.
SHAN G C, ZHAO H B.Detection and Analysis of Ship Radiated-Noise Line Spectrum[J]. Ship Electronic Engineering, 2014, 34(10): 119-122.