| 摘要: |
| 现有的雷达信号分选方法多采用静态分选机制，通过批处理方式对接收到的全部脉冲数据进行分选。然而，这些批处理方式无法挖掘雷达信号的动态变化特性，且在高脉冲密度场景下处理效率显著下降。针对上述问题，提出一种基于均值漂移与多维云模型的雷达信号分选方法。该方法首先引入均值漂移聚类算法对累积脉冲数据进行静态分选，并采用多维云模型对分选结果的参数特征进行建模，以实现对雷达信息的有效表征；然后，通过计算后续脉冲与各云模型之间的隶属度，完成对新脉冲的分选与归属判断，同时能够实现雷达信息的更新与演化；最后，基于云模型之间的相似度度量实现相似雷达的融合。实验结果表明：与现有分选方法相比，该方法可准确捕捉雷达的动态演化特性，在强干扰环境下仍能保持95%以上的分选正确率，并在脉冲样本数密集的情况下具备良好的处理效率。 |
| 关键词: 在线分选 数据流 均值漂移 云模型 |
| DOI： |
| 分类号:TN957.51 |
| 基金项目:国家自然科学基金(62071238)，江苏省自然科学基金(BK20191399)，江苏省研究生科研与实践创新计划项目(KYCX25_1645) |
|
| A Radar Signal Sorting Method Using Mean Shift and Multi-dimensional Cloud Model |
|
|
|
|
| Abstract: |
| Existing radar signal sorting methods mostly rely on static sorting mechanisms, they perform sorting on the entire set of received pulses through batch processing. However, these batch-processing approaches cannot capture the dynamic variation characteristics of radar signals and suffer from significantly reduced efficiency in high pulse density scenarios. To address these issues, a radar signal sorting method based on mean shift and multidimensional cloud model is proposed. Firstly, the mean shift clustering algorithm is introduced to perform static sorting on accumulated pulse data, and a multidimensional cloud model is employed to characterize the parameter features of the sorted results to effectively represent radar information. Then, by computing the membership degree between subsequent pulses and each cloud model, sorting and attribution of new pulses are accomplished, while radar information can be updated and evolved. Finally, similar radars are merged based on the similarity measurement between cloud models. Experimental results demonstrate that, compared with existing sorting methods, the proposed approach accurately captures the dynamic evolution of radar signals, maintains over 95% sorting accuracy under strong interference, and achieves high processing efficiency even in dense pulse scenarios. |
| Key words: online sorting data stream mean shift cloud model |