基于轨迹段DBSCAN的船舶轨迹聚类算法  

Ship Trajectory Clustering Algorithm Based on DBSCAN

作　　者：江玉玲[1] 熊振南[2] 唐基宏[1] JIANG Yuling;XIONG Zhennan;TANG Jihong(Chengyi College, Jimei University, Xiamen 361021, China;Navigation College, Jimei University, Xiamen 361021, China)

机构地区：[1]集美大学诚毅学院,厦门361021 [2]集美大学航海学院,厦门361021

出　　处：《中国航海》2019年第3期1-5,共5页Navigation of China

基　　金：福建省教育厅中青年教师教育科研项目(JAT170912);福建省自然科学基金(2016J01243)

摘　　要：船舶自动识别系统(Automatic Identification System, AIS)数据中蕴藏着大量的海上交通特征,为挖掘AIS数据中有关船舶运动规律有效的、潜在的信息,提出一种改进型轨迹段DBSCAN(Density-Based Spatitcal Clustering of Applications with Noise)的聚类算法。船位转向角和航速变化量作为信息度量对船舶轨迹进行分段,采用离散Frechet距离作为轨迹相似度度量,利用类似DBSCAN算法对轨迹段进行聚类,得出船舶运动典型轨迹。以天津港为例,采用改进的轨迹段DBSCAN算法对船舶轨迹进行聚类,能从一定程度上提高聚类的效果和准确率,为进一步研究船舶异常行为打下基础。An improved DBSCAN (Density-Based Spatial Clustering of Applications with Noise) trajectory clustering is developed to study ship AIS (Automatic Identification System) data collected from ship broadcasts and to learn the law of ship motion. The ship trajectories are segmented according to the ship course / speed variations. According to the similarity measured by the discrete Frechet distance, the trajectory subsections are clustered with the DBSCAN algorithm, and the typical trajectory of ship motion is determined. The improved DBSCAN algorithm is used to study the ship trajectories in Tianjin Port for illustration. It can improve the clustering effect and accuracy to a certain extent, and lay a foundation for further research on ship abnormal behavior.

关 键 词：船舶轨迹 分段 相似度度量 DBSCAN 轨迹聚类 

分 类 号：U675.7[交通运输工程—船舶及航道工程]