摘要
把核方法引入探地雷达目标特征提取中,先对雷达特征数据进行基于核的非线性变换投影到高维空间;然后在高维空间应用传统的探地雷达目标特征提取方法(LDA方法和PCA方法)提取雷达数据特征。对PCA、LDA、基于核方法的PCA和基于核方法的LDA四种探地雷达目标特征提取方法进行对比研究,对实测数据处理的结果表明:无监督的学习方法不适合探地雷达目标特征提取,所提基于核方法的LDA探地雷达目标特征提取方法效果最好。并详细阐述了基于核方法的LDA探地雷达目标特征提取方法。
The kernel method is proposed to extract target feature of the ground penetrating radar (TFGPR). Mapped GPR data to the high dimension space by the kernel method, then extracted the TFGPR in the high dimension space by the conventional TFGPR extraction method (i. e. LDA and PCA). The treatment results of measurement data show that the unsupervised methods are not fit for the TFGPR extraction, and the proposed LDA method based on the kernel super to the others.
出处
《电波科学学报》
EI
CSCD
北大核心
2005年第5期671-674,共4页
Chinese Journal of Radio Science
关键词
探地雷达
特征提取
核方法
ground penetrating radar (GPR), feature extraction, kernel method