针对船舶航行轨迹数据因时间序列特征导致的不确定性，难以准确捕捉航行轨迹数据的关键信息，影响最终检测精度的问题，提出深度学习算法下船舶航行轨迹异常行为检测方法。依据船舶自动识别系统（AIS）获取船舶航行轨迹数据并实施预处理后，提取AIS船舶航行轨迹的时间序列特征。将AIS船舶航行轨迹特征输入门控循环单元循环神经网络深度学习模型进行处理，依据更新门、重置门处理，最终实现船舶航行轨迹异常行为检测。实验证明，该方法在不同种类船舶类型混合情况下，检测偏差均小于1.1%，时效性均能够保证在97.5%以上，能够实时监控船舶航行状态，预警异常行为，降低船舶发生事故风险，推动海上交通管理智能化发展。
Aiming at the uncertainty of ship navigation trajectory data caused by time series characteristics, which makes it difficult to accurately capture key information of navigation trajectory data and affects the final detection accuracy, a deep learning algorithm based method for detecting abnormal behavior of ship navigation trajectory is proposed. Based on the Automatic Identification System (AIS), the AIS ship navigation trajectory data is obtained and preprocessed to extract the time series features of the AIS ship navigation trajectory. Input the AIS ship navigation trajectory features into the gate controlled recurrent unit recurrent neural network deep learning model for processing, and based on the update gate and reset gate processing, ultimately achieve abnormal behavior detection of ship navigation trajectory. Experimental results have shown that this method has a detection deviation of less than 1.1% and a timeliness of over 97.5% in mixed situations of different types of ships. It can monitor the navigation status of ships in real time, warn of abnormal behavior, reduce the risk of ship accidents, and promote the intelligent development of maritime traffic management.
2025,47(15): 164-167 收稿日期：2025-4-7
DOI：10.3404/j.issn.1672-7649.2025.15.027
分类号：U675
作者简介：冀娜(1989-),女,高级工程师,研究方向为人工智能和图形处理
参考文献：
[1] 张阳, 高曙, 何伟, 等. 基于多模型融合的内河船舶航行轨迹预测方法[J]. 中国机械工程, 2022, 33(10): 1142-1152.
[2] 焦建芳, 包端华, 胡正中. 基于预设性能的自适应神经网络船舶轨迹跟踪[J]. 华中科技大学学报(自然科学版), 2022, 50(4): 77-82.
[3] 李高才, 张新宇, 蒋晨星, 等. 海港航道水域船舶异常行为检测[J]. 大连海事大学学报, 2024, 50(4): 31-40+78.
[4] 黄谊文, 胡勤友, 董思诺. 桥区水域船舶异常行为检测[J]. 上海海事大学学报, 2023, 44(2): 1-5+82.
[5] 彭耀武, 陈辰, 刘敬贤, 等. 基于BO-GRU和AKDE的船舶异常行为识别[J]. 中国航海, 2024, 47(3): 10-20.
[6] 陈家豪, 刘钊, 张明阳, 等. 融合主成分分析与支持向量机的船舶异常行为识别方法[J]. 中国航海, 2023, 46(3): 126-134.
[7] 朱木清, 邹欢. 深度学习下动态目标识别算法优化仿真[J]. 计算机仿真, 2023, 40(12): 321-324+336.
ZHU M Q, ZOU H. Optimization and simulation of dynamic target recognition algorithm with deep learning[J]. Computer Simulation, 2023, 40(12): 321-324+336.
[8] 胡银杰, 齐乐, 沈德翔. 基于AIS的船舶进出港轨迹数据压缩研究[J]. 武汉理工大学学报, 2024, 46(3): 116-124.
