海量电磁数据中雷达信号的高效分选方法

针对海量电磁数据中雷达信号难以进行快速准确分选的问题,提出一种新的聚类分选方法,即改进k-means算法的Map Reduce并行化实现方法。通过引入初始聚类中心个数k1、最大聚类中心个数kmax和距离门限rt3个参数,克服了k-means算法需要事先确定k值和易受孤立点影响的局限;基于Hadoop平台实现了对改进k-means算法的Map Reduce并行化,克服了k-means算法串行实现时间复杂度高的局限。最后,实验表明改进k-means算法取得了更高的分选准确率,Map Reduce并行化后具有良好的加速比和扩展性,能够很好地对海量电磁数据中雷达信号进行高效分选。

An Efficient Sorting Method of Radar Signals in the Massive Electromagnetism Data

Aimed at the problem that it’s hard to sort the radar signals in the massive electromagnetism data quickly and accurately,a new clustering algorithm is proposed,that is improved k-means algorithm using MapReduce programming mode. The initial clustering centers number k1,the maximum clustering centers number kmax and the distance threshold rt are introduced by the improved k-means clustering algorithm,to overcome the confines of the k-means clustering algorithm that it needs the pre-determined k value and is prone to be effected by isolated individual data point. Based on the Hadoop platform,Parallel implementation of the improved k-means clustering algorithm using MapReduce programming mode is realized,to overcome the confines of the k-means algorithm that serial implementation has a high time complexity. Finally, the experimental result validates the improved k -means clustering algorithm has high sorting precision, and shows the parallel implementation of the improved k-means clustering algorithm using MapReduce programming mode owns good speedup ratio and scalability. It also can sort the radar signals in the massive electromagnetism data efficiently.