船舶AIS轨迹快速自适应谱聚类算法  

Fast self-tuning spectral clustering algorithm for AIS ship trajectory

作　　者：牟军敏[1,2] 陈鹏飞 贺益雄[1,2] 张行健 朱剑峰[3] 荣昊 MOU Junmin;CHEN Pengfei;HE Yixiong;ZHANG Xingjian;ZHU Jianfeng(School of Navigation, Wuhan University of Technology, Wuhan 430063, China;Hubei Key Laboratory of Inland Shipping Technology, Wuhan 430063, China;Shenzhen Shekou Zhaogang Passenger Transportation Industry Co. Ltd., Shenzhen 518067, China;Centre for Marine Technology and Ocean Engineering (CENTEC), Instituto Superior Técnico, Universidade de Lisboa, Lisbon 1049-001 , Portugal)

机构地区：[1]武汉理工大学航运学院,湖北武汉430063 [2]湖北省内河航运重点实验室,湖北武汉430063 [3]深圳招商蛇口国际邮轮母港有限公司,广东深圳518067 [4]斯本科技大学海洋技术与工程中心,里斯本1049-001

出　　处：《哈尔滨工程大学学报》2018年第3期428-432,共5页Journal of Harbin Engineering University

基　　金：国家自然科学基金面上项目(51579201);武汉理工大学研究生自主创新基金项目(2015-zy-109, 175212005)

摘　　要：为了对船舶AIS轨迹数据进行快速聚类,本文提出了一种基于Hausdorff距离的船舶轨迹快速自适应谱聚类算法(fast self-tune spectral clustering,FSSC)。在保留轨迹特征的情况下,利用Douglas-Peucker(DP)算法对船舶轨迹数据进行预处理;基于Hausdorff距离,设计自动选取尺度参数的相似度度量函数,构造相似度矩阵并采用谱聚类算法对船舶轨迹进行聚类。以长江口水域船舶实际AIS数据为样本对算法进行了验证,结果表明:聚类结果能够准确提取水域船舶主要航路,算法消耗系统资源少,计算速度快。该方法对水域船舶主要航路识别,提高海事监管效率等方面具有参考意义。To conduct fast clustering of automatic identification system(AIS)ship trajectory data,in this paper,we propose a fast self-tuning spectral clustering(FSSC)algorithm based on the Hausdorff distance.The trajectory data are pre-processed by the Douglas-Peucker(DP)algorithm,which preserves the trajectory characteristics.Based on the Hausdorff distance,trajectory similarity measurement function and similarity matrix that can automatically choose the scaling parameters are proposed,and a spectral clustering algorithm is used to cluster the ship trajectory.To verify the proposed method,we selected the estuary of the Yangtze River as a case study and the results indicate that the FSSC can obtain the main route in the marine navigation area.The consumption of computer resources is small,and the calculation speed is much faster than the usual clustering method.The proposed algorithm can provide a reference for the identification of main ship routes and improve the efficiency of maritime traffic management.

关 键 词：船舶自动识别系统 船舶轨迹 DOUGLAS-PEUCKER算法 数据压缩 HAUSDORFF距离 谱聚类 

分 类 号：U675.7[交通运输工程—船舶及航道工程]