| 摘要: |
| 针对大擦地角条件下海杂波中目标特征检测性能提升的需求，本文开展了机载平台雷达对海探测试验，获取了擦地角从25°变化至75°条件下海面与目标实测回波数据；通过提取与分析时域、频域等不同变换域特征，利用Fisher Score与Pearson相关系数筛选出了高区分度和低冗余度的特征子集（TEM、MS、RVE），并对比了自回归（AR）、支持向量回归（SVR）及卡尔曼滤波（KF）三种模型对特征随角度变化的序列预测性能；实验结果表明，相比于AR与SVR，卡尔曼滤波表现出最优的预测精度。本文研究揭示了海杂波与目标回波特征的随擦地角的变化规律，验证了擦地角变化条件下特征序列的可预测性，为大擦地角条件下雷达目标检测方法设计提供支撑。 |
| 关键词: 大擦地角 海杂波 特征提取 序列预测 |
| DOI： |
| 分类号:TN959 |
| 基金项目:国家自然科学基金项目（面上项目，重点项目，重大项目） |
|
| Predictive Modeling of Radar Sea Clutter and Target Echo Features Under High Grazing Angles |
|
|
|
|
| Abstract: |
| To improve the performance of radar target detection in sea clutter under high grazing angles, this paper conducts a sea prob experiment using an airborne radar, and acquires the measured echo data of sea surface and targets with grazing angles ranging from 25° to 75°. By extracting and analyzing features in different transform domains such as time domain and frequency domain, a feature subset (TEM、MS、RVE) characterized by high discriminability and low redundancy is selected using Fisher Score and Pearson correlation coefficients. Furthermore, this study evaluates the performance of three models, Autoregressive (AR), Support Vector Regression (SVR), and Kalman Filter (KF), in predicting feature sequences as they evolve with the grazing angle. Experimental results show that the Kalman Filter achieves superior prediction accuracy compared with AR and SVR. This study reveals the variation patterns of sea clutter and target features with grazing angles and validates the predictability of feature sequences under varying geometry, providing a theoretical support for designing radar target detection methods under high grazing angles. |
| Key words: high grazing angle sea clutter feature extraction sequence prediction |