系统工程与电子技术 ›› 2023, Vol. 45 ›› Issue (12): 3726-3733.doi: 10.12305/j.issn.1001-506X.2023.12.02

• 电子技术 • 上一篇    

地面目标HRRP识别的稳健性特征选择方法

刘丹阳1,*, 吴堃2, 朱永锋1, 张永杰2, 周剑雄1   

  1. 1. 国防科技大学电子科学学院, 湖南 长沙 410073
    2. 北京遥感设备研究所, 北京 100854
  • 收稿日期:2022-08-09 出版日期:2023-11-25 发布日期:2023-12-05
  • 通讯作者: 刘丹阳
  • 作者简介:刘丹阳 (1998—), 女, 硕士研究生, 主要研究方向为高分辨雷达成像、目标特性分析与识别
    吴堃 (1986—), 男, 高级工程师, 硕士, 主要研究方向为雷达系统、雷达信号处理
    朱永锋 (1979—), 男, 副研究员, 博士, 主要研究方向为雷达波形设计、雷达信号处理、雷达抗干扰与目标识别
    张永杰 (1980—), 男, 研究员, 博士, 主要研究方向为高分辨雷达系统与目标检测识别
    周剑雄 (1977—), 女, 教授, 博士, 主要研究方向为高分辨雷达成像、目标特性分析与识别

Robust feature selection method for ground target HRRP recognition

Danyang LIU1,*, Kun WU2, Yongfeng ZHU1, Yongjie ZHANG2, Jianxiong ZHOU1   

  1. 1. College of Electronics Science and Technology, National University of Defense Technology, Changsha 410073, China
    2. Beijing Institute of Remote Sensing Device, Beijing 100854, China
  • Received:2022-08-09 Online:2023-11-25 Published:2023-12-05
  • Contact: Danyang LIU

摘要:

特征选择是雷达高分辨一维距离像目标识别的关键步骤, 可降低特征维度, 提高特征稳健性。提出一种基于散度的特征选择方法, 采用该方法对适用于距离像地面目标识别的特征集合进行特征选择, 得到优选的特征子集后再进入分类器网络进行识别。采用地面目标仿真数据和实测数据进行神经网络分类器识别实验。实验结果表明: 在距离像信噪比、俯仰角和距离分辨力参数变化的情况下, 基于散度的特征选择方法在基本保持或提升特征集的识别性能的前提下, 能保持甚至提升识别的稳健性, 具有较好的应用价值。

关键词: 雷达目标识别, 高分辨一维距离像, 特征提取, 特征选择, 地面目标

Abstract:

Feature selection is a key step in radar high resolution range profile (HRRP) target recognition, which can reduce the features dimension and improve the features robustness. A feature selection method based on divergence is proposed. This method is used to select feature subset from the feature set which is suitable for the recognition of ground targets range profile. Then the well-selected feature subset is sent to the classifier network for recognition. The ground target simulation data and filed data are used to conduct neural network classifier recognition experiments. The experimental results show that when the signal to noise ratio, elevation angle, and resolution of range profile change, the feature selection method based on divergence can not only maintain or improve the recognition ability of the feature set, but also maintain or even improve the robustness of the recognition, which has good practical value.

Key words: radar target recognition, high resolution range profile, feature extraction, feature selection, ground target

中图分类号: