| 摘要: |
| K分布杂波参数估计在海洋雷达自适应目标检测中起着关键作用。传统的矩估计器通过联立两个矩求解参数，其估计性能受限于有限矩的信息。因此，本文提出一种基于多维矩特征联合的参数估计方法，旨在拓展矩信息的维度。首先，从观测数据中，提取多个精心设计的线性矩和对数矩，构建一个特征向量。其次，将传统基于统计分布的参数估计问题转换为非线性优化问题。然后，通过引入梯度提升决策树（Gradient Boosting Decision Tree, GBDT）算法，建立特征向量和形状参数之间的函数关系，实现形状参数的估计。此外，推导证明特征向量与尺度参数的独立性以及二阶矩只依赖于尺度参数，从而解决两个参数估计的相关性问题。最后，仿真和实测数据结果表明，所提估计器能利用多个矩的丰富信息，进一步提高参数估计性能。特别是在小形状参数时，其估计性能显著优于现有矩估计法和zrlog（z）期望法。 |
| 关键词: 海杂波 K分布 参数估计 多维矩特征 |
| DOI：DOI:10.3969/j.issn.1672-2337.2024.01.011 |
| 分类号:TN957.51 |
| 基金项目:国家自然科学基金（No.62201184）；江苏省高等学校基础科学（自然科学）研究项目（No.21KJB510036） |
|
| Parameter Estimation of K⁃Distributed Clutter via Multidimensional Moment Feature Combination |
|
SHI Sainan， GAO Jijuan， LI Dongchen
|
|
1. Shool of Electronic and Information Engineering, Nanjing University of Information Science & Technology, Nanjing 210044, China;2. System Engineering Research Institute, China State Shipbuilding Corporation, Beijing 100094, China
|
| Abstract: |
| Parameter estimation of K?distributed clutter plays an important role in adaptive target detection for marine radars. The traditional moment estimator solves the parameters by combining two moments and its performance is limited by the information of finite moments. Therefore, a parameter estimation method via multidimensional moment feature combination is proposed to expand the dimension of moment information. Firstly, multiple well?designed linear moments and logarithmic moments are extracted from the observed data to construct a feature vector. Secondly, the traditional parameter estimation problem based on statistical distribution is transformed into a nonlinear optimization problem. Thirdly, the gradient boosting decision tree （GBDT） algorithm is introduced to establish the functional relationship between the feature vector and the shape parameter to achieve the estimation of the shape parameter. Besides, it is proved that the feature vector is independent of the scale parameter and the second moment only depends on the scale parameter, so as to solve the relationship of the two parameters in estimation. Finally, the simulation and measured data results show that the proposed estimator can take advantage of the rich information of multiple moments to further improve the performance of parameter estimation. Especially for small shape parameters, its estimation performance is significantly better than the existing moment estimation methods and zrlog（z） expectation method. |
| Key words: sea clutter K⁃distribution parameter estimation multidimensional moment features |
