摘要
在雷达地面目标识别中,采用合适的特征选择算法有助于在复杂的特征空间中挑选出对不同类型目标具有较强区分能力的特征。结合信息论和合作博弈理论,提出一种基于Owen值的特征选择算法,旨在选择出与类别相关度高、特征与特征之间冗余性低、依赖性强的最优特征子集,达到在雷达目标识别中提高识别率的目的。利用雷达实测数据验证其性能,结果表明,所选特征子集的平均识别准确率优于两种经典的Filter式特征选择算法,且该算法具有良好的噪声/杂波稳健性。
In radar ground target recognition,suitable feature selection algorithm contributes to selecting the features with high discrimination in the complex characteristic space for different types of target.On the basis of information theory and cooperative game theory,a novel feature selection algorithm is proposed based on the Owen value.The proposed feature selection algorithm can select the optimal feature subset,in which the features are of high relevance with its class,and have low redundancy and strong dependence with each other,thus can greatly improve the recognition performance of radar target.Experiments conducted on the real data show that:1) The average recognition rate of the proposed method is better than that using the two classical Filter algorithms;and 2) The proposed feature selection algorithm is robust against noise/clutter.
作者
朱红茹
刘峥
王晶晶
马晓瑛
ZHU Hongru;LIU Zheng;WANG Jingjing;MA Xiaoying(National Laboratory of Radar Signal Processing,Xidian University,Xi'an 710071,China)
出处
《电光与控制》
CSCD
北大核心
2020年第11期6-9,58,共5页
Electronics Optics & Control
基金
国家自然科学基金(61301282)。
关键词
雷达目标识别
特征选择
合作博弈
Owen值
radar target recognition
feature selection
cooperative game
Owen value
