基于主成分分析和机器学习的激光水下尾流识别

王旭阳; 陈伟

doi:10.3724/CO.2026-0127

基于主成分分析和机器学习的激光水下尾流识别

doi: 10.3724/CO.2026-0127

cstr: 32171.14.CO.2026-0127

王旭阳,
陈伟^,

国防科技大学气象海洋学院, 湖南长沙 410073

基金项目: 无

详细信息

作者简介:
王旭阳（2002—），男，江苏南京人，硕士，主要从事海洋光学方面的研究。E-mail: xuyangwang24@126.com 联系方式：18013015727

陈伟（1983—），男，山东肥城人，博士，副教授，硕士生导师，主要从事海洋探测技术和海洋光学方面的研究。E-mail：kevinkobegames@126.com 联系方式：13975860480

中图分类号: P733.3
计量
- 文章访问数: 7
- HTML全文浏览量: 3
- PDF下载量: 0
- 被引次数: 0
出版历程
- 网络出版日期: 2026-10-22

Laser recognition of underwater wakes based on principle component analysis and machine learning

WANG Xu-yang,
CHEN Wei^,

College of Meteorology and Oceanography, National University of Defense Technology, Changsha, Hunan 410037, China

Funds: xxxx

More Information

Corresponding author: kevinkobegames@126.com

摘要

摘要:
为了实现对水下尾流的识别，本研究结合主成分分析（PCA）和不同的机器学习方法，使用激光前向传输信号特性对不同类型尾流进行识别。在处理前向传输的光信号时，首先计算时域和频域特征，再使用主成分分析对原始特征进行降维，最后通过机器学习处理降维后的特征，实现对水槽条件下不同水下推进器尾流的识别。同时，湖上试验也验证了该方法在实际应用中的可行性。结果表明，在水槽条件下，表现最优的机器学习方法为支持向量机（SVM），综合准确率为89.93%。而在湖上试验中，K近邻（KNN）和随机森林（RF）的识别性能更为优秀，综合准确率分别为96.39%和96.25%。随着尾流强度的改变和推进器类型的增加，不同尾流的差异程度逐渐减小，PCA特征边界出现混淆。此外，信号的时域特征和频域特征对不同识别任务的重要性具有显著差异。实验研究和湖上试验结果均说明该方法可以有效实现对不同水下尾流的识别。
- 激光探测 /
- 尾流 /
- 机器学习 /
- 前向散射
Abstract:
To achieve the recognition of underwater wakes, this study integrates principal component analysis (PCA) with various machine learning methods to classify different types of wakes using the characteristics of forward-propagated laser signals. In processing the forward-propagated optical signals, time-domain and frequency-domain features are first computed, followed by dimensionality reduction of the original features via PCA. The reduced features are then fed into machine learning classifiers to distinguish wakes generated by different underwater thrusters under tank conditions. Furthermore, lake test validates the practical feasibility of the proposed method. Results indicate that under tank conditions, the optimal machine learning method is support vector machine (SVM), achieving a accuracy of 89.93%. Whereas in the lake test, the optimal methods are k-nearest neighbors (KNN) and random forest (RF), with accuracies of 96.39% and 96.25% respectively. As the wake intensity varies and the number of thruster types increases, the inter-wake dissimilarity decreases, and the PCA feature boundaries become blurred. Additionally, the importance of time-domain versus frequency-domain features differs significantly across classification tasks. Both experimental and lake test results demonstrate that the proposed method can effectively discriminate among different underwater wakes.
- laser detection /
- wakes /
- machine learning /
- forward scattering

HTML全文

图 1 数据处理流程图

Figure 1. Data processing flowchart

下载: 全尺寸图片幻灯片

图 2 实验平台

Figure 2. Experimental platform

下载: 全尺寸图片幻灯片

图 3 探测光路示意图

Figure 3. Schematic of the detection light path

下载: 全尺寸图片幻灯片

图 4 水下推进器

Figure 4. Underwater propellers

下载: 全尺寸图片幻灯片

图 5 五叶桨强尾流和异常尾流的光信号

Figure 5. Optical signal of 5-Blade intense and abnormal wake

下载: 全尺寸图片幻灯片

图 6 不同条件下各方法的识别结果。（a）不同强度的四叶桨尾流；

（b）不同推进器类型的中强度尾流；（c）不同推进器类型和强度的尾流

Figure 6. The classification results of methods under different conditions.

(a) wakes with different intensities generated by 4-blade propeller; (b) wakes with moderate intensity generated by different propellers; (c) wakes with different intensities generated by different propellers

下载: 全尺寸图片幻灯片

图 7 不同条件下的PCA特征空间散点图。（a）不同强度的四叶桨尾流；

（b）不同推进器类型的中强度尾流；（c）不同推进器类型和强度的尾流

Figure 7. PCA feature space scatter plot under different conditions.

下载: 全尺寸图片幻灯片

图 8 仅使用时域特征和频域特征时的识别结果。（a）不同强度的四叶桨尾流；

（b）不同推进器类型的中强度尾流；（c）不同推进器类型和强度的尾流

Figure 8. Classification results using time-domain and frequency-domain features respectively.

下载: 全尺寸图片幻灯片

图 9 湖上试验的目标船只和设备

Figure 9. Detection targets and device of lake test

下载: 全尺寸图片幻灯片

图 10 归一化的平均频域曲线

Figure 10. Normalized average frequency curve

下载: 全尺寸图片幻灯片

表 1 实验数据的时域特征

Table 1. Time-domain features of experimental data

时域特征类型	Ⅰ类	Ⅱ类
特征名称	均值	标准差
	最大值	偏度
	最小值	峰度
	范围	均方根
	第一四分位数	绝对中位差
	中位数	四分位距
	第三四分位数	\

下载: 导出CSV

表 2 使用不同时域特征的识别结果

Table 2. Classification results using different time-domain features

机器学习方法	时域特征识别准确率
机器学习方法	Ⅰ类	Ⅱ类	Ⅰ类 + Ⅱ类
RF	84.63%	86.41%	87.48%
NB	74.19%	73.70%	77.22&
KNN	85.11%	84.81%	86.59%
SVM	84.70%	88.26%	89.93%
注：识别时同时使用了频域特征。

下载: 导出CSV

表 3 湖上试验测试结果

Table 3. Classification results of lake test

机器学习方法	目标1 识别准确率	目标2 识别准确率	综合准确率
RF	97.50%	95.00%	96.25%
NB	95.00%	92.50%	93.89%
KNN	95.50%	97.50%	96.39%
SVM	85.50%	95.00%	90.14%

下载: 导出CSV

参考文献(18)

[1]	张银波, 李思宁, 姜鹏, 等. PCA特征提取和弹性BP神经网络的水下气泡识别[J]. 红外与激光工程, 2021, 50(6): 209-215. doi: 10.3788/IRLA20200352 ZHANG Y B, LI S N, JIANG P, et al. Underwater bubbles recognition based on PCA feature extraction and elastic BP neural network[J]. Infrared and Laser Engineering, 2021, 50(6): 209-215. (in Chinese). doi: 10.3788/IRLA20200352
[2]	田恒斗, 金良安, 王涌, 等. 考虑单气泡运动特性的舰船尾流气泡分布研究[J]. 兵工学报, 2011, 32(9): 1126-1130. TIAN H D, JIN L A, WANG Y, et al. Study on wake bubble distribution based on single bubble motion characteristics[J]. Acta Armamentarii, 2011, 32(9): 1126-1130. (in Chinese).
[3]	宗思光, 张鑫, 曹静, 等. 舰船尾流激光探测跟踪方法与试验[J]. 红外与激光工程, 2023, 52(3): 202205. doi: 10.3788/IRLA20220507 ZONG S G, ZHANG X, CAO J, et al. Method and experiment of laser detection and tracking of ship wake[J]. Infrared and Laser Engineering, 2023, 52(3): 202205. (in Chinese). doi: 10.3788/IRLA20220507
[4]	梁善勇, 王江安, 宗思光, 等. 基于多重散射强度和偏振特征的舰船尾流气泡激光探测方法[J]. 物理学报, 2013, 62(6): 060704. LIANG SH Y, WANG J A, ZONG S G, et al. Laser detection method of ship wake bubbles based on multiple scattering intensity and polarization characteristics[J]. Acta Physica Sinica, 2013, 62(6): 060704. (in Chinese).
[5]	宗思光, 段子科, 张鑫, 等. 基于同步累积法的舰船尾流微弱信号检测方法研究[J]. 中国光学(中英文), 2024, 17(6): 1272-1280. doi: 10.37188/CO.2024-0014 ZONG S G, DUAN Z K, ZHANG X, et al. Detection method of weak ship wake signals based on the synchronous accumulation method[J]. Chinese Optics, 2024, 17(6): 1272-1280. (in Chinese). doi: 10.37188/CO.2024-0014
[6]	刘罡, 李永胜, 刘礼文, 等. 基于时频分析和迁移学习的舰船尾流检测方法[J]. 水下无人系统学报, 2022, 30(4): 465-473. doi: 10.11993/j.issn.2096-3920.202108013 LIU G, LI Y SH, LIU L W, et al. Method of ship wake detection based on time-frequency analysis and transfer learning[J]. Journal of Unmanned Undersea Systems, 2022, 30(4): 465-473. (in Chinese). doi: 10.11993/j.issn.2096-3920.202108013
[7]	吴鹏飞, 胡田田, 王姣, 等. 拉盖尔-高斯涡旋光束经混合海洋介质相位屏传输的闪烁因子[J]. 激光与光电子学进展, 2024, 61(15): 1501002. doi: 10.3788/LOP231750 WU P F, HU T T, WANG J, et al. Scintillation index of Laguerre-Gaussian vortex beam propagation through an ocean mixed-medium phase screen[J]. Laser & Optoelectronics Progress, 2024, 61(15): 1501002. (in Chinese). doi: 10.3788/LOP231750
[8]	卢腾飞, 刘永欣, 吴志军. 涡旋光束在不同盐度的水下湍流中的传输特性的实验研究(英文)[J]. 中国光学, 2022, 15(1): 111-118. LU T F, LIU Y X, WU ZH J. Experimental investigation on propagation characteristics of vortex beams in underwater turbulence with different salinity[J]. Chinese Optics, 2022, 15(1): 111-118.
[9]	KHAMEES H T, ALGBURI S, KOTB H E. Computation of the scintillation index and BER of super Lorentz Gaussian laser beam in a slant propagation at atmospheric turbulence[J]. Ain Shams Engineering Journal, 2025, 16(3): 103295. doi: 10.1016/j.asej.2025.103295
[10]	项玉龙, 陈伟. 气泡散射影响下的水下光传输特性研究[J]. 红外与激光工程, 2024, 53(12): 202403. doi: 10.3788/IRLA20240321 XIANG Y L, CHEN W. Research on underwater light transmission characteristics under influence of bubbles scattering[J]. Infrared and Laser Engineering, 2024, 53(12): 202403. (in Chinese). doi: 10.3788/IRLA20240321
[11]	邱琳琳, 朱卫纲, 李永刚, 等. 基于深度学习的SAR图像目标检测综述[J]. 兵工自动化, 2026, 45(3): 72-76. doi: 10.7690/bgzdh.2026.03.013 QIU L L, ZHU W G, LI Y G, et al. Survey of SAR image target detection based on deep learning[J]. Ordnance Industry Automation, 2026, 45(3): 72-76. (in Chinese). doi: 10.7690/bgzdh.2026.03.013
[12]	杨长生, 苟文博, 梁红. 基于一维卷积神经网络的舰船尾流检测[J]. 水下无人系统学报, 2023, 31(6): 839-846. doi: 10.11993/j.issn.2096-3920.2022-0052 YANG CH SH, GOU W B, LIANG H. Ship wake detection based on one-dimensional convolutional neural network[J]. Journal of Unmanned Undersea Systems, 2023, 31(6): 839-846. (in Chinese). doi: 10.11993/j.issn.2096-3920.2022-0052
[13]	王振, 樊超阳, 杨苏辉, 等. 独立成分分析方法在水下激光雷达中的应用[J]. 光学学报, 2024, 44(18): 1801002. doi: 10.3788/AOS231965 WANG ZH, FAN CH Y, YANG S H, et al. Application of independent component analysis method in underwater lidar[J]. Acta Optica Sinica, 2024, 44(18): 1801002. (in Chinese). doi: 10.3788/AOS231965
[14]	WANG X Y, CHEN W. Classification of underwater propeller wakes based on machine learning[J]. Journal of Physics: Conference Series, 2026, 3178: 012035. doi: 10.1088/1742-6596/3178/1/012035
[15]	唐勐, 张宇. 激光探测尾流微气泡的偏振特性研究[J]. 红外与激光工程, 2020, 49(1): 010500. doi: 10.3788/IRLA202049.0105006 TANG M, ZHANG Y. Research on polarization characteristics of micro-bubbles laser detection in wake[J]. Infrared and Laser Engineering, 2020, 49(1): 010500. (in Chinese). doi: 10.3788/IRLA202049.0105006
[16]	宗思光, 张鑫, 杨劭鹏, 等. 舰船尾流气泡目标激光后向散射特性研究[J]. 中国光学(中英文), 2023, 16(6): 1333-1342. doi: 10.37188/CO.2023-0043 ZONG S G, ZHANG X, YANG SH P, et al. Laser backscattering characteristics of ship wake bubble target[J]. Chinese Optics, 2023, 16(6): 1333-1342. (in Chinese). doi: 10.37188/CO.2023-0043
[17]	焦贵金, 张建生, 闫林波, 等. 尾流气泡幕对激光的传输特性研究[J]. 水下无人系统学报, 2024, 32(5): 833-838. doi: 10.11993/j.issn.2096-3920.2023-0093 JIAO G J, ZHANG J SH, YAN L B, et al. Laser transmission characteristics of wake bubble curtain[J]. Journal of Unmanned Undersea Systems, 2024, 32(5): 833-838. (in Chinese). doi: 10.11993/j.issn.2096-3920.2023-0093
[18]	JAMALI M V, KHORRAMSHAHI P, TASHAKORI A, et al. Statistical distribution of intensity fluctuations for underwater wireless optical channels in the presence of air bubbles[C]. 2016 Iran Workshop on Communication and Information Theory (IWCIT), IEEE, 2016: 1-6.