基于DBSCAN算法的船舶轨迹自适应层次聚类  

Adaptive Hierarchical Clustering of Ship Trajectory with DBSCAN Algorithm

作　　者：赵梁滨 史国友[1,2] 杨家轩 ZHAO Liangbin;SHI Guoyou;YANG Jiaxuan(Navigation College, Dalian Maritime University, Dalian 116026, China ;The Key Laboratory of Navigation Safety Guarantee Liaoning Province, Dalian 116026, China)

机构地区：[1]大连海事大学航海学院,辽宁大连116026 [2]辽宁省航海安全保障重点实验室,辽宁大连116026

出　　处：《中国航海》2018年第3期53-58,共6页Navigation of China

基　　金：国家自然科学基金(51579025);国家高技术研究发展计划(863计划)课题2009AA045003;辽宁省自然科学基金(201602084)

摘　　要：为更好地从船舶自动识别系统(Automatic Indentification System,AIS)数据中挖掘信息,科学地感知水上交通态势,针对聚类算法(Density-Based Spatial Clustering of Applications with Noise,DBSCAN)在水上交通情景中的参数选取问题,提出一种基于DBSCAN的船舶轨迹自适应层次聚类方法。通过分析DBSCAN算法的特性,根据数据集内在分布规律及拟聚类效果的变化规律来确定参数;结合统计学理论进行层次聚类,来适应密度分布不均的船舶轨迹数据。以琼州海峡船舶轨迹数据为例,运用VC软件和MATLAB软件进行验证。验证结果表明:该方法能够在大量复杂的船舶轨迹中发现具有相似性的轨迹群,且结果与实际交通流相符,可为航道建设及海事监管等提供辅助决策。An adaptive hierarchical clustering method based on DBSCAN algorithm is proposed for extracting information from AIS data and understanding the traffic situation on water. The behavior of DBSCAN algorithm in the situation of water transportation is analyzed and a few measures are taken to improve the clustering,including defining features through analyzing the distribution law of the data and the variation law of clustering results so as to eliminate the possible error caused by the data density difference between trajectories. The model is implemented with VC and MATLAB. A case study is carried out with the ship trajectory data from Qiongzhou strait,which indicates that the method is able to correctly recognize trajectories of similar characteristics from complicated traffic data. The method can be a useful tool for traffic flow investigation.

关 键 词：交通运输工程 船舶轨迹聚类 统计分析 聚类算法 船舶自动识别系统 Inverse Gaussian 

分 类 号：U675.81[交通运输工程—船舶及航道工程] U665.261[交通运输工程—船舶与海洋工程]