基于复合MeanShift聚类算法的雷达信号分选

打开文本图片集

中图分类号： TN971+.1 文献标志码：A

Abstract： A radar signal sorting method based on the composite Mean Shift clustering algorithm is proposed，aiming to solve the challenges in radar signal processing in complex electromagnetic environments. Since the traditional time-of-arrival （TOA）-based sorting method is difficult to cope with the intertwining of signals from multiple sources，the method to identify and classify diffrent types of radar signals by pre-sorting them through the multi-dimensional data of pulse descriptors is proposed. Gaussian kernel function is introduced into the Mean Shift algorithm to weight the data points and improve the accuracy of the algorithm，and the adaptive bandwidth strategy is proposed to improve the clustering flexibility and accuracy by dynamically adjusting the bandwidth.Then for the influence of noise in the practical application environment，the DBSCAN algorithm is combined with the denoising process，which further improves the accuracy and stability of clustering.Verified by simulation experiments，the composite Mean Shift + DBSCAN algorithm shows strong anti-interference ability and high clustering accuracy under high pulse density and noise interference.Overall，the research provides new ideas and methods for eficient sorting of radar signals in complex electromagnetic environments，which can better meet the needs of radar signal processing in modern battlefields.

Key Words： pulse description word; radar signal sorting;machine learning；clustering algorithm; Mean Shift clustering

0 引言

当前陆战场电磁环境复杂，信号源种类多样，雷达信号样式趋于复杂化，各种形式雷达脉冲信号（如参差雷达、频率捷变雷达）混杂，单方面针对雷达信号到达时间（TOA）的分选无法很好地满足战场侦察的需求，并且在实际应用中可基于脉冲描述字（PDW）的多维数据对雷达信号进行分选，而采集外部信号生成的脉冲描述字基本为多源信号，复杂信号交织在一起易导致PDW参数混乱，无法直接对其应用常用的基于TOA的雷达分选算法，如序列差直方图法[2。（剩余7290字）