基于改进孤立森林的海面小目标检测方法  

Detection Method of Sea Surface Small Target Based on Improved Isolation Forest

作　　者：李雨笑 胡居荣[1] 邢延潇 戴天石 张伟杰 LI Yuxiao;HU Jurong;XING Yanxiao;DAI Tianshi;ZHANG Weijie(Hohai University,Nanjing 210024,China)

机构地区：[1]河海大学,江苏南京210024

出　　处：《雷达科学与技术》2024年第6期628-636,共9页Radar Science and Technology

基　　金：国家自然科学基金(No.62271190)。

摘　　要：针对海杂波背景下的雷达弱目标检测问题以及孤立森林(isolation Forest, iForest)算法在处理高维数据时未充分利用雷达回波信号特征信息的问题,提出了一种基于改进多特征联合的孤立森林弱目标检测方法。该方法通过分析实测海杂波数据在时域、频域和时频域的特性构建了丰富的高维特征矩阵,在iForest算法中融合主成分分析算法进行数据降维,引入平均相关度构成双参数降维准则,以平衡主成分与原始特征之间的相关性。仿真结果表明,所提改进方法在不同海况以及极化方式下均能够有效提升海杂波背景下雷达弱目标检测的性能,且在虚警概率较低的情况下仍有较高的检测概率。To address the problem of radar small target detection under sea clutter conditions and the issue thatthe isolation forest(iForest)algorithm does not fully utilize radar echo signal features when dealing with high‑dimensionaldata, a method for weak target detection based on improved multi‑feature joint isolation forest is proposed. This methodconstructs a rich high‑dimensional feature matrix by analyzing the characteristics of actual sea clutter data in time domain,frequency domain and time‑frequency domain. The principal component analysis is integrated into the isolationforest algorithm for data dimensionality reduction, introducing an average correlation‑based dual‑parameter criterion fordimensionality reduction to balance the correlation between the principal components and the original features.The simulationresults demonstrate that the proposed method effectively enhances the detection performance of radar weak targetunder sea clutter conditions across different sea states and polarization modes, while maintaining a high detection probabilityunder low false alarm rates.

关 键 词：雷达目标检测 孤立森林算法 主成分分析算法 多特征联合 海杂波 

分 类 号：TN957.51[电子电信—信号与信息处理]