基于VMD与PNN的一维雷测数据快速分析系统

Rapid Analysis System for Radar Measurement Data Based on VMD and PNN

摘要:

针对一维雷测数据中广泛高频振荡与局部不规则振荡耦合导致的时域边界模糊、机理判断笼统与分析效率低下等问题，采用变分模态分解（VMD）与概率神经网络（PNN）构建快速分析系统。引入麻雀搜索算法优化VMD输入参数，利用VMD窄带分量特性实现不规则振荡自适应切分，从时域、频域、相关性等多维度提取29项特征，构建训练样本集并采用PNN完成振荡机理分类。经实测数据集验证，系统对方位、俯仰测角数据的振荡模式基本分类准确率分别为93.9%与94.1%，单次处理5000点数据耗时3至5秒，满足雷测数据快速分析的工程需求。

Abstract:

To address the challenges of temporal boundary ambiguity, imprecise mechanism identification, and low analytical efficiency arising from the coupling of pervasive high-frequency oscillations and localized irregular oscillations in one-dimensional radar tracking measurement data, a rapid analysis system is developed based on Variational Mode Decomposition (VMD) and Probabilistic Neural Networks (PNN). A Sparrow Search Algorithm (SSA) is introduced to optimize the input parameters of VMD, and the narrowband component properties of VMD are leveraged to achieve adaptive segmentation of irregular oscillations. A total of 31 features are extracted from multi-dimensional perspectives encompassing the time domain, frequency domain, and correlation analysis, upon which a training sample set is constructed and PNN is employed to accomplish oscillation mechanism classification. Validation against measured datasets demonstrates that the system achieves oscillation pattern classification accuracies of 93.9% and 94.1% for azimuth and elevation angle measurement data, respectively, with a single processing cycle for 5,000 data points completed within 3–5 seconds, thereby satisfying the engineering requirements of rapid analysis for radar tracking measurement data.