基于数据库的收缩型航迹聚类模型仿真研究

Deflating Track Clustering Model Simulation Based on Database Research

针对传统聚类算法只能选取少量数据源进行仿真分析,得到的聚类效果不能真实体现数据流的宏观特征的问题,考虑了海量航迹数据中的离群点检测以及离群点剔除,提出一种新的基于数据库的收缩型航迹聚类仿真模型,将三维空间网格化,建立K-means聚类和层次聚类双重交互算法,对网格中的离群点进行识别并剔除。解决了航迹聚类中的关键技术问题。通过对西北地区3 G海量二次雷达数据的聚类仿真分析,使航迹聚类仿真的耗时从h级降低至s级,并且得到的航迹分布特征清晰,验证了新模型对于海量数据宏观特征提取具有可行性和优越性,模型和算法对全国二次雷达航迹数据仿真具有借鉴意义。

In view of the traditional clustering algorithm can only choose a small amount of data source to go on simulation analysis, the clustering effect cannot really reflect the data flow of macroscopic character-istics of the problem. In this paper, considering the vast track detecting outliers in the data and eliminate outliers,put forward a new simulation model based on database of deflating track clustering, the three - di-mensional space grids,the establishment of K means clustering and hierarchical clustering algorithm, double interaction to identify and eliminate outliers in grid, Solve the key technical problems in the track is clustering. The clustering of simulation analysis to the northwest of 3 G huge amounts of secondary radar data, make track clustering simulation time consuming levels from hours reduced to seconds, and the track distribution characteristics are clear,macro feature extraction for huge amounts of data to verify the new model has the feasibility and superiority, this model and algorithm have the significance of reference to the secondary radar track data simulation.