﻿ 稠密度聚类在舰船网络微弱信号自适应增强中的应用
 舰船科学技术  2022, Vol. 44 Issue (5): 138-141    DOI: 10.3404/j.issn.1672-7649.2022.05.029 PDF

1. 金华市现代制造与材料高新技术研发中心，浙江 东阳 322100;
2. 浙江广厦建设职业技术大学，浙江 东阳 322100

Application of density-clustering in adaptive enhancement of weak signals in ship networks
YU Xiao-xia1,2, PAN Guang-yong1,2
1. Jinhua Modern Manufacturing and Material High-tech RESEARCH and development Center, Dongyang 322100, China;
2. Zhejiang Guangsha Construction Vocational and Technical University, Dongyang 322100, China
Abstract: To improve the signal and overall communication quality of ship network, an adaptive enhancement method of ship network weak signal based on dense clustering is proposed. The dense density clustering algorithm is used to detect the weak signal of ship network. The weak signal is decomposed by empirical mode decomposition method, and the signal enhancement of ship network weak signal is realized by combining the noise removed by wavelet transform method and singular spectrum analysis method. The results show that this method can effectively decompose the detected weak signal and denoise the components. After denoising, each signal component not only retains the original detailed characteristics, but also effectively eliminates the interference noise. The reconstructed enhanced signal is clear as a whole, which can effectively ensure the overall communication quality of the ship network.
Key words: density clustering     ship network     weak signal     adaptive enhancement     empirical mode decomposition     singular spectrum analysis
0 引　言

1 稠密度聚类的舰船网络微弱信号自适应增强方法 1.1 基于稠密度聚类的舰船网络微弱信号检测

 ${\boldsymbol{l}}\left( {x,y} \right) = \left[ \begin{gathered} 0\mathop {}\nolimits_{} l\left( {0,1} \right)\mathop {}\nolimits_{} l\left( {0,2} \right)\mathop {}\nolimits_{} \cdots l\left( {0,m - 1} \right) \hfill \\ \mathop {}\nolimits_{} \mathop {}\nolimits_{} 0\mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\limits_{} l\left( {1,2} \right)\mathop {}\nolimits_{} \cdots l\left( {1,m - 1} \right) \hfill \\ \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \vdots \mathop {}\nolimits_{} \mathop {}\nolimits_{} \hfill \\ \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} 0\mathop {}\nolimits_{} l\left( {m - 2,m - 1} \right) \hfill \\ \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} \mathop {}\nolimits_{} 0\mathop {}\nolimits_{} \mathop {}\nolimits_{} \end{gathered} \right]。$ (1)

 ${r_0} = \min \left\{ {l\left( {b,a} \right),N} \right\}，$ (2)

 ${r_M} = \left\{ \begin{gathered} {r_0},M = 1 ，\\ {r_{M - 1}} + 2{r_0}/\left( {1 + M} \right),M \geqslant 2 。\end{gathered} \right.$ (3)

 $\rho = \frac{{\left| {{D_i}} \right|}}{{\left| B \right|}} 。$ (4)

 $\bar \rho = \frac{1}{m}\sum\limits_{i = 1}^m {\rho \left( {{D_i}} \right)}。$ (5)

 图 1 稠密度聚类的舰船网络微弱信号整体检测过程图 Fig. 1 Overall detection process diagram of weak signals in ship network based on density clustering
1.2 EMD-SSA的舰船网络微弱信号自适应增强 1.2.1 基于EMD的舰船网络微弱信号分解

 $a = d + c ，$ (6)

 $a = \sum\limits_{i = 1}^J {IM{F_i} + {f_J}}。$ (7)

 $S = \arg \max \left[ {{F_p}} \right],1 \leqslant p \leqslant J。$ (8)

 ${F_p} \cong \frac{1}{J}\sum\limits_{j = 1}^J {{{\left[ {IM{F_p}} \right]}^2}}，$ (9)

 $g = \sum\limits_{i = 1}^S {IM{F_i}}，$ (10)

 $h = \sum\limits_{i = s}^J {IM{F_i}} 。$ (11)

1.2.2 小波变换的舰船网络微弱信号噪声IMF分量去噪

1）选取恰当的小波基函数，对微弱信号的噪声IMF分量g实施小波变换，获取到一组小波系数，以σj,k表示；

2）依据相应规则设置阈值函数W，采用阈值函数W处理不同尺度上的小波系数，将其中比该阈值函数小的小波系数置为0，并收缩或者保存比该阈值函数大的小波系数，获得预估小波系数σ′j,k

3）运用预估小波系数σj,k实施小波重构，获得预估IMF分量g′，此即为小波去噪后的微弱信号部分IMF分量。

1.2.3 SSA的舰船网络微弱信号IMF分量去噪

 图 2 奇异谱创建三角形 Fig. 2 Singular spectra create triangles

 ${a'^2} = {b'^2} + {c'^2} - 2b'c'\cos A'。$ (12)

 $\cos C' ={{{a'}^2} + {{b'}^2} - {{c'}^2}}/{2a'b'}，$ (13)

 $h' = \sum\limits_{w = 1}^q {{\delta ^w}{H^w}} ,1 \leqslant w \leqslant T 。$ (14)

 $\hat a = {f_J} + h' + g'。$ (15)
2 应用结果分析

 图 3 实验舰船网络微弱信号稠密度聚类检测结果 Fig. 3 Clustering detection results of weak signal density of experimental ship network

 图 4 舰船网络初始微弱信号 Fig. 4 Initial weak signal in ship network

 图 5 舰船网络微弱信号a的增强信号 Fig. 5 Enhancement signal of weak signal A in ship network
3 结　语

 [1] 时培明, 袁丹真, 张文跃, 等. 基于时延反馈多稳随机共振的微弱信号检测方法[J]. 计量学报, 2020, 41(7): 868-872. [2] 于希明, 洪硕, 于金祥, 等. 可见光遥感图像船舶目标数据增强方法研究[J]. 仪器仪表学报, 2020, 41(11): 261-269. [3] 刘刚, 行鸿彦, 张金玉. 一种优化的微弱信号检测方法[J]. 现代雷达, 2019, 41(6): 23-30. [4] 王骞, 何培宇, 徐自励. 利用奇异谱分析的深度神经网络语音增强方法[J]. 信号处理, 2020, 36(6): 902-910. [5] 张鑫涛, 马福民, 曹杰, 等. 基于混合度量与类簇自适应调整的粗糙模糊K-means聚类算法[J]. 模式识别与人工智能, 2019, 32(12): 1141-1150. [6] 黄钢, 瞿伟斌, 许卉莹. 基于改进密度聚类算法的交通事故地点聚类研究[J]. 交通运输系统工程与信息, 2020, 20(5): 169-176. [7] 王飞宇, 胡志祥, 黄潇. 基于密度峰值聚类算法的模态参数识别[J]. 振动与冲击, 2019, 38(2): 172-178. [8] 崔彩霞, 曹付元, 梁吉业. 基于密度峰值聚类的自适应欠采样方法[J]. 模式识别与人工智能, 2020, 33(9): 811-819. [9] 徐丽丽, 董一鸿, 王雄, 等. 基于K-sup稠密子图的大规模复杂网络概要算法及可视化[J]. 计算机辅助设计与图形学学报, 2019, 31(3): 400-411. [10] 杨璐, 李醒飞, 纪越, 等. 基于EMD-DFA-NLM的电涡流传感器信号去噪方法[J]. 传感技术学报, 2019, 32(9): 1318-1323. DOI:10.3969/j.issn.1004-1699.2019.09.007 [11] 罗红梅, 宋维琪, 邢漪冉, 等. 基于改进经验模态分解的地震弱信号增强处理方法[J]. 地球物理学进展, 2019, 34(1): 167-173. DOI:10.6038/pg2019BB0402 [12] 李敏, 刘岩, 马然, 等. 一种用于海水DOC微弱信号去噪处理的小波多阈值算法研究[J]. 传感技术学报, 2021, 34(1): 75-79. DOI:10.3969/j.issn.1004-1699.2021.01.012 [13] 徐博, 刘斌, 刘德政, 等. 基于小波理论的舰船IMU加速度计信号处理算法[J]. 中国舰船研究, 2020, 15(2): 151-158+164.