自适应高斯混合模型海上移动对象浮标轨迹聚类研究  

Research on Floating Cluster Trajectory Clustering of Moving Object in Adaptive Gaussian Mixture Model

作　　者：葛丽阁 孙伟[1] GE Li-ge;SUN Wei(College of Information Engineering,Shanghai Maritime University,Shanghai 201306)

机构地区：[1]上海海事大学信息工程学院,上海201306

出　　处：《现代计算机（中旬刊）》2018年第8期3-8,共6页Modern Computer

基　　金：国家自然基金青年基金项目(No.61203240)

摘　　要：海洋轨迹具有震荡性大和非受限性等特征而成为研究的热点和难点。针对现有算法研究的不足,提出一种自适应高斯混合模型的聚类算法,可在很大程度上可以避免人工设定聚类个数的烦扰,首先通过设定一个较小的起始聚类簇个数作为EM(即期望最大化算法)的初始化参数,然后设定一个合适的阈值不断迭代来确定是否增加聚类个数,从而得到最优的聚类簇。通过实验仿真表明自适应高斯混合模型聚类算法不仅实用性和可靠性较高,与传统的高斯混合模型和HMM聚类相比可信度较高,而且该算法适用于海上浮标非受限复杂轨迹,通过对轨迹的聚类分析可以为未来的工作,例如海上搜救、轨迹异常点检测和海洋运输业的繁荣,提供重要的意义。Due to the characteristics of large and unconstrained oceanic trajectories have become the hot and difficult research. The clustering algorithm of adaptive Gaussian mixture model can avoid the annoyance of the number of clustering manually. First, by setting a smaller number of initial clustering clusters as the initialization parameters of EM. Then setting a suitable threshold to iterate determining whether to increase the number of clusters, so as to obtain the optimal cluster. The experimental results show that the adaptive Gaussian mixture model clustering algorithm is not only practical and reliable, but also has high reliability compared with the traditional Gaussian mixture model clustering, and the algorithm is suitable for the unrestricted complex trajectory of the sea buoy, clustering analysis of trajectories can provide important implications for future work such as maritime search and rescue, trajectory anomaly detection and marine transport boom.

关 键 词：高斯混合模型 自适应聚类 浮标轨迹 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]