海量电磁数据中雷达信号的高效分选方法

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

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。

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 Map Reduce programming mode. The initial clustering centers number k1,the maximum clustering centers number kmaxand the distance threshold rtare 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 Map Reduce 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 Map Reduce programming mode owns good speedup ratio and scalability. It also can sort the radar signals in the massive electromagnetism data efficiently.