基于改进DBSCAN的船舶轨迹聚类方法研究

Research on Clustering Method of Vessel Trajectory Based on Improved DBSCAN

为了辅助海上安全监管,科学分析海上交通安全态势,根据有着大量船舶航行特征的AIS数据,提出一种改进轨迹段的DBSCAN算法进行轨迹聚类。在对原始AIS数据进行预处理后,结合航向变化率和航速变化率获取特征点的方式来进行轨迹分段,采用融合距离(MD)作为船舶轨迹的距离计算方法,并且充分考虑航向信息和航速信息来进行相似度度量,传统的DBSCAN算法只对点进行聚类,改进后的DBSCAN算法可以对轨迹分段后的轨迹子段进行聚类分析,通过实验分析,可以得到船舶典型运动轨迹,实验对比结果显示,论文所提聚类方法在一定程度上可以获得更好的聚类效果。船舶轨迹聚类是船舶轨迹预测的基础,因此得到更好的聚类结果有利于提高后续预测的准确度。

In order to assist maritime safety supervision and scientific analysis of maritime traffic safety situation,a DBSCAN algorithm with improved trajectory segments is proposed for trajectory clustering based on AIS data with a large number of vessel navigation characteristics. After pre-processing the raw AIS data,the trajectory segmentation is performed by combining the heading change rate and speed change rate to obtain feature points. The Merge Distance(MD)is used as the distance calculation method for vessel trajectory,and the heading information and speed information are fully considered for the similarity measure. The traditional DBSCAN algorithm clusters only the points,the improved DBSCAN algorithm can cluster and analyze the sub-segments of the trajectory after the segmentation of the trajectory. Through the experimental analysis,the typical motion trajectory of the vessel can be obtained,and the experimental comparison results show that the clustering method proposed in this paper can obtain a better clustering effect to a certain extent. Vessel trajectory clustering is the basis of vessel trajectory prediction,so getting better clustering results is helpful to improve the accuracy of subsequent prediction.