一种改进的数据场聚类算法

An improved data field clustering algorithm

雷达信号分选是电子侦察中的关键步骤,针对传统聚类算法需要先验知识、算法需要人为设定参数、对孤立噪声点敏感和对初始聚类中心的选取对聚类效果有直接的影响、容易出现"增批"缺点,提出一种改进的数据场聚类算法。该算法计算所有的数据对象的势值,通过寻找势心来确定初始聚类中心和聚类数目,根据数据对象的势值大小和阈值进行比较,剔除孤立噪声点,将数据对象划分到距离最近的聚类中心的那一类中完成聚类。文中仿真了12部雷达信号,包括了常规雷达、抖动雷达、参差雷达和捷变频雷达,雷达参数相近或交叠。仿真结果表明,改进的数据场聚类算法有良好的聚类效果。

Radar signal sorting is a key step in electronic reconnaissance.Aiming at the shortcomings of traditional clustering algorithms,for example,it requires a priori knowledge,it needs to set parameters artificially for the algorithm,it is sensitive to isolated noise points,and the selection of initial clustering center has a direct impact on the clustering effect,it is prone to have"increased batching",and so on,an improved data field clustering algorithm is proposed.By calculating the potential values of all data objects,the algorithm determines the initial cluster center and the number of clusters by finding the potential center.According to the potential value of the data object and the threshold value,the isolated noise points are eliminated,and the data objects are divided into the nearest ones.Clustering is done in that class of clustering centers.In this paper,12 radar signals are simulated,including conventional radar,jitter radar,staggered radar and agile radar.The radar parameters are similar or overlapping.The simulation results show that the improved data field clustering algorithm has good clustering effect.