船舶自动识别系统（AIS）对于保障船舶航行安全与海事监管意义重大。详细分析了AIS系统面临的信号干扰与遮挡、电磁干扰、算法局限性以及抗干扰能力弱等问题，提出基于卷积神经网络（CNN）的优化方法，涵盖从数据收集、预处理到模型构建与训练的完整流程，通过交叉熵损失函数、Adam优化算法及学习率退火策略，有效提升模型性能。对比实验结果表明：基于CNN优化的AIS系统在不同场景下信号准确率显著提高，不同天气条件下虚警率明显降低，提高船舶航行安全及海事监管的效率与可靠性。
Automatic Ship Identification System (AIS) is of great significance to ensure the safety of ship navigation and maritime supervision. The problems faced by AIS system, such as signal interference and occlusion, electromagnetic interference, algorithm limitations and weak anti-interference ability, are analyzed in detail. An optimization method based on convolutional neural network (CNN) is proposed, covering the complete process from data collection and preprocessing to model construction and training. Through cross-entropy loss function, Adam optimization algorithm and learning rate annealing strategy, the AIS system is designed to solve the following problems: Effectively improve model performance. The comparative experimental results show that the signal accuracy of the AIS system optimized by CNN is significantly improved under different scenarios, and the false alarm rate is significantly reduced under different weather conditions, which improves the efficiency and reliability of ship navigation safety and maritime supervision.
2025,47(7): 174-178 收稿日期：2024-9-16
DOI：10.3404/j.issn.1672-7649.2025.07.032
分类号：U667.65
基金项目：海南热带海洋学院崖州湾创新研究院重大科技计划项目（2023CXYZD001）
作者简介：吴淑婷(1985-),女,硕士,讲师,研究方向为软件工程及数据分析
参考文献：
[1] 叶萍. 基于卷积神经网络的船舶自动识别系统[J]. 船电技术, 2021, 41(3): 61-64.
YE P. Ship automatic identification system based on convolutional neural network[J]. Marine Electric & Electronic Technology, 2021, 41(3): 61-64.
[2] 曲颖丽. 基于卷积神经网络的船舶识别[D]. 大连：大连海事大学, 2020.
[3] 熊先华, 宾进宽, 姜飞. 基于卷积神经网络的过闸船舶越限识别方式[J]. 西部交通科技, 2024(5): 186-189.
XIONG X H, BIN J K, JIANG F. Over - limit recognition method of ships passing through the lock based on convolutional neural network[J]. Western China Communications Science & Technology, 2024(5): 186-189.
[4] 程静, 王荣杰, 曾光淼, 等. 基于更快区域卷积神经网络的多视角船舶识别[J]. 哈尔滨工程大学学报, 2023, 44(10): 1832-1840.
CHENG J, WANG R J, ZENG G M, et al. Multi - view ship recognition based on faster region convolutional neural network[J]. Journal of Harbin Engineering University, 2023, 44(10): 1832-1840.
[5] 孙祎芸, 樊臻, 董山玲, 等. 基于双流对称特征融合网络模型的海洋船舶目标识别[J]. 控制理论与应用, 2022, 39(11): 2009-2018.
SUN Y Y, FAN Z, DONG S L, et al. Ocean ship target recognition based on dual - stream symmetric feature fusion network model[J]. Control Theory & Applications, 2022, 39(11): 2009-2018.
[6] 陈会伟, 刘树美, 刘培学, 等. 超尺度自导注意力网络的遥感船舶识别[J]. 计算机工程, 2021, 47(10): 314-320.
CHEN H W, LIU S M, LIU P X, et al. Remote sensing ship recognition of super - scale self - guided attention network[J]. Computer Engineering, 2021, 47(10): 314-320.
[7] 李兆桐, 孙浩云. 基于全卷积神经网络的船舶检测和船牌识别系统[J]. 计算机与现代化, 2019(12): 72-77.
LI Z T, SUN H Y. Ship detection and ship plate recognition system based on fully convolutional neural network[J]. Computer and Modernization, 2019(12): 72-77.
[8] 闫河, 王鹏, 董莺艳, 等. 一种CNN与ELM相结合的船舶分类识别方法[J]. 重庆理工大学学报(自然科学), 2019, 33(1): 53-57.
YAN H, WANG P, DONG Y Y, et al. A ship classification and recognition method combining CNN and ELM[J]. Journal of Chongqing University of Technology (Natural Science), 2019, 33(1): 53-57.
