摘要
基于目标高分辨率距离像的雷达自动目标识别技术在军事和民用上都有巨大的应用价值。但是由于雷达目标高分辨距离像的姿态敏感性以及高特征维数,造成了其非线性可分性。针对此问题,本文提出了一种基于最大间隔核优化的雷达目标高分辨距离像识别方法。本方法首先采用了最大间隔准则算法来优化数据依赖核函数,然后利用支持向量机分类器实现了雷达目标高分辨距离像识别,最后进行了基于5种战斗机目标高分辨距离像的实验仿真。实验结果表明了基于最大间隔核优化的目标识别算法对于SVM分类器可以有效实现核函数优化,从而能够提高目标识别性能。
The radar automatic target recognition （RATR）base on the radar target High-Resolution Range Profile （HRRP）plays an important role in military and civilian field.But the target aspect sensitivity and high feature dimensionality of the target HRRP cause the nonlinear separability of the HRRP.Aiming at this problem,this paper proposes a new radar target HRRP recognition method based on the maximal margin kernel optimization.At first,the maximal margin principle is used for the data-dependent kernel optimization,and then the Support Vector Machine （SVM）is applied for the radar target HRRP recognition as a classifier.At last,the recognition simulation experiment based on 5 kinds of aircraft targets HRRP has been done.And the experimental results show that the proposed method based on the maximal margin kernel optimization can effectively optimize the kernel parameters considering SVMclassifier,and thus improve the recognition performance.
出处
《信号处理》
CSCD
北大核心
2014年第7期783-788,共6页
Journal of Signal Processing
基金
国家自然科学基金(61001150)
2013年江西省教育厅科技项目(DB201304184)
关键词
雷达目标高分辨距离像
支持向量机
核函数优化
雷达目标识别
最大间隔准则
radar target high-resolution range profile
support vector machine
kernel optimization
radar target recogni-tion
maximal margin principle
