| 摘要: |
| 针对极化合成孔径雷达（PolSAR）舰船目标检测中复杂海况下欧氏线性处理存在的检测性能下降问题，提出了一种基于对数流形特征的扰动抑制与舰船目标检测方法。首先，利用对数欧几里得映射（LEM）将极化协方差矩阵（PCM）从弯曲的黎曼流形投影至平直切空间，通过对数运算的压缩特性抑制重尾分布并克服欧氏膨胀效应；其次，提取切空间中独立且完备的9维实值以增强表征能力；在此基础上，分别针对目标PCM已知与未知两种典型场景，构建了对数流形最优极化检测器（LE-OPD）与对数流形极化白化检测器（LE-PWD）。仿真与实测数据表明，所提方法在Wishart、K 分布、G0分布及真实混合杂波背景下均能保持优异的检测性能。 |
| 关键词: 极化合成孔径雷达 舰船检测 对数欧几里得映射 黎曼流形 |
| DOI： |
| 分类号:TN957.52 |
| 基金项目:国家自然科学基金(62501625),国家资助博士后研究人员计划(GZC20252839) |
|
| Ship Target Detection in PolSAR Images Based on Log-Euclidean Manifold Features |
|
|
|
|
| Abstract: |
| To address the performance degradation issue caused by Euclidean linear processing in Polarimetric Synthetic Aperture Radar (PolSAR) ship target detection under complex sea conditions, a novel PolSAR disturbance suppression and ship target detection method based on Log-Euclidean manifold features is proposed. First, Log-Euclidean Mapping (LEM) is used to project the Polarimetric Covariance Matrix (PCM) from a curved Riemannian manifold to a flat tangent space, where the compression property of the logarithmic operation suppresses heavy-tailed distributions and overcomes the Euclidean expansion effect. Second, a set of independent and complete 9-dimensional real-valued features is extracted from the tangent space to enhance representation capability. On this basis, for the two typical scenarios where the target PCM is known or unknown, the Log-Euclidean Optimal Polarimetric Detector (LE-OPD) and the Log-Euclidean Polarimetric Whitening Detector (LE-PWD) are constructed, respectively. Simulation and real data experiments demonstrate that the proposed method maintains excellent detection performance under Wishart, K-distribution, G0-distribution, and real mixed clutter backgrounds. |
| Key words: polarimetric synthetic aperture radar ship detection log-Euclidean mapping Riemannian manifold |