摘要
传统船舶交通异常识别方法在大雾天气环境下,存在挖掘算法对船舶轨迹异常状态辨识度降低的问题。通过分析发现,原因在于传统方法中没有引入大雾天气对船舶轨迹检测信号的扰动变量,导致轨迹检测数据与挖掘算法之间出现数据断链,降低了数据挖掘的识别效果。因此,提出大雾天气海上船舶交通异常挖掘识别方法分析。首先通过LSTM算法,将大雾天气扰动特征代入挖掘神经网络,获得带有大雾扰动特征神经网络;接着,根据大雾扰动特征建立混合高斯船舶轨迹模型,为交通异常识别提供基础数据;然后,通过Spark分布式挖掘算法,完成对船舶交通异常数据的挖掘识别。通过仿真实验,对传统挖掘识别方法与提出方法效果进行多组数据对比,证明提出挖掘识别方法的有效性。
In the fog weather environment, the traditional ship traffic anomaly identification method has the problem of reducing the identification degree of ship track anomaly state by mining algorithm. Through the analysis, it is found that the disturbance variable of the ship trajectory detection signal is not introduced in the traditional method, which leads to the data break chain between the trajectory detection data and the mining algorithm, and reduces the recognition effect of the data mining. Therefore, the identification method of marine traffic anomaly mining in fog weather is proposed. First, by LSTM algorithm, the disturbance characteristics of fog weather are brought into the mining neural network to obtain the characteristic neural network with fog disturbance, and then according to the disturbance characteristics of fog A hybrid Gauss ship trajectory model is established to provide basic data for traffic anomaly identification;then, by Spark a distributed mining algorithm, the mining and identification of ship traffic anomaly data is completed;through simulation experiments, The effectiveness of proposed mining identification method is proved by comparing the traditional mining identification method with the proposed method.
作者
王正星
朱玉柱
王春瑶
WANG Zheng-xing;ZHU Yu-zhu;WANG Chun-yao(Dalian Maritime University,Dalian 116026,China)
出处
《舰船科学技术》
北大核心
2021年第2期49-51,共3页
Ship Science and Technology
关键词
大雾天气
船舶
交通异常
挖掘识别
heavy fog weather
ships
traffic anomalies
excavation identification
