多步骤船舶轨迹聚类方法研究与实现  

Research and Implementation of Multi-step Ship Track Clustering Method

作　　者：蒋通 崔良中[1] 周钢[1] 张社国 JIANG Tong;CUI Liangzhong;ZHOU Gang;ZHANG Sheguo(College of Electronic Engineering,University of Naval Engineering,Wuhan 430033;No.92682 Troops of PLA,Zhanjiang 524000)

机构地区：[1]海军工程大学电子工程学院,武汉430033 [2]中国人民解放军92682部队,湛江524000

出　　处：《舰船电子工程》2021年第9期53-57,92,共6页Ship Electronic Engineering

摘　　要：船舶自动识别系统(AIS)数据中包含着船舶大量行为特征,如何挖掘潜藏在其中的有价值信息一直是该领域研究的重难点之一,轨迹聚类是目前研究海上目标行为模式、发现海上交通规律的一种有效方法。论文针对现有对AIS船舶轨迹聚类过程不够系统完整的问题,对AIS数据进行预处理后,提出一种多步骤船舶轨迹聚类方法,主要包括轨迹压缩、轨迹点匹配与距离相似性度量和轨迹聚类等三部分。选取长江入海口河段部分水域一定时间内的真实AIS数据进行实例验证,所得结果不仅能找出该水域的主要航道,而且还发现了异常轨迹的存在。该方法对于AIS船舶轨迹聚类的系统研究有一定的参考价值。The data of ship automatic identification system(AIS)contains a large number of ship behavior characteristics.How to excavate the valuable information hidden in them has always been one of the important and difficult points in this field.Track clustering is currently an effective method to study the behavior pattern of marine target and discover the law of maritime traffic.This paper presents a method,which mainly includes three parts,which are trajectory compression,distance and trajectory clustering.Based on the real AIS data which belong to some waters of the Yangtze River estuary for a certain time,the results can not only find out the main channel of the water area,but also find the existence of abnormal trajectory.This method has certain reference value for the systematic study of AIS ship trajectory clustering.

关 键 词：船舶轨迹 多步骤聚类 相似性度量 轨迹点匹配 DBSCAN 

分 类 号：U675.7[交通运输工程—船舶及航道工程] TP391[交通运输工程—船舶与海洋工程]