2021年世界交通运输大会水运学部会议 基于改进K中心点聚类的船舶典型轨迹自适应挖掘算法

Meeting of the Waterborne Transport Division, World Transport Convention 2021 (WTC 2021)Adaptive algorithm for ship typical trajectory mining based on improved K-medoids clustering

针对目前船舶典型轨迹的挖掘多以轨迹段作为基本单元,导致聚类对象较为复杂且聚类参数难以确定的问题,本文提出一种基于改进K中心点聚类的船舶典型轨迹自适应挖掘算法。算法以轨迹点作为聚类对象,分析船舶的航速、航向特征并对轨迹点进行压缩;将分段均方根误差引入K中心点聚类算法,实现聚类参数的自适应选择;提取其中的聚类中心点作为轨迹特征点,得到不同类别船舶的典型轨迹。以天津港主航道船舶自动识别系统(automatic identification system,AIS)数据为例,基于地理信息系统平台ArcGIS实现聚类结果的可视化展示。实验结果表明,运用该算法得到的船舶典型轨迹与实际相符,自适应程度较高。研究结果对于辅助船舶轨迹异常检测及挖掘海上交通特征具有重要意义。

Currently,the trajectory segment is taken as the basic unit in the ship typical trajectory mining,which leads to complex clustering objects and difficulty in determining cluster parameters.To solve the problem,the adaptive algorithm for ship typical trajectory mining based on the improved K-medoids clustering is proposed.The algorithm takes trajectory points as clustering objects.The characteristics of ship speed and course are analyzed,and the trajectory points are compressed;the segmented root mean square error is introduced into the K-medoids clustering algorithm to realize the adaptive selection of clustering parameters;the cluster center points are extracted as the trajectory feature points,and the typical trajectories of different types of ships are obtained.Taking automatic identification system(AIS)data of the main channel of Tianjin Port as an example,this paper realizes the visualization of clustering results based on the geographic information system platform ArcGIS.The experimental results show that the ship typical trajectories obtained by the algorithm are consistent with the actual situation,and the degree of self-adaptation is high.The research results are of great significance to assist the detection of abnormal ship trajectories and mine the characteristics of marine traffic flow.