基于Spark的船舶航行轨迹聚类方法

Clustering Method of Ship's Navigation Trajectory Set Based on Spark

依托船舶自动识别系统(Automatic Identification System,AIS)数据,利用云计算并结合聚类算法,对船舶历史数据进行轨迹聚类分析,构建船舶航行正常轨迹模型,为实时检测船舶异常轨迹奠定基础,进而为提高水上交通监管智能化水平提供新方法。针对目前轨迹聚类算法效率低等问题,基于Spark内存计算技术及数据分区思想,提出一种改进的并行子轨迹聚类算法SPDBSCANST(Parallel DBSCAN of Sub Trajectory Based on Spark)。以长江航道武汉段船舶航行数据为例进行试验验证,并通过可视化方式呈现。结果表明,改进后的算法的聚类效率和效果都有明显提升。

Constructing normal navigation trajectory model through processing historical AIS（ Automatic Identification System） data of ships with the trajectory clustering algorithm is a way of setting up the reference for real-time detection of abnormal ships trajectory. Aimed at the problem of low efficiency of the current trajectory clustering algorithm,an improved parallel sub trajectory clustering algorithm is proposed named as SPDBSCANST（ Parallel DBSCAN of Sub Trajectory Based on Spark） featuring Spark memory computing technology and data partition. The algorithm is verified with the ship navigation data of Yangtze River Waterway. The visualization of the trajectories is also achieved. The experiments show that the efficiency of the improved clustering algorithm is increased significantly.