径向基函数神经网络在超宽带探地雷达目标材质识别中的应用  

The Application of RBF Neural Network in Material Recognition of Ultra-Wideband Ground-Penetrating Radar

作　　者：郑军庭[1] 李建[2] 李建勋[1] 

机构地区：[1]上海交通大学信息与控制研究所,上海200030 [2]河南司法警官职业学院,郑州450002

出　　处：《上海交通大学学报》2006年第1期98-102,共5页Journal of Shanghai Jiaotong University

基　　金：上海市科技发展基金资助项目(015115038)

摘　　要：基于探地雷达回波信号进行处理以识别地下埋设的目标,始终是困扰雷达应用的关键,根据数据时间轴截断抑制直达波,利用宽相关处理进行信号滤波和典型数据自动提取,提高回波信号的信噪比.针对提取的典型道数据运用Welch功率谱处理,得到的特征数据归一化处理后作为径向基函数神经网络的输入,实现对地下埋设目标材质的自动识别与分类.在此基础上,分析了不同截断点对目标材质识别结果的影响.实测数据处理表明,本方法可以有效地实现对Fe、Al与土壤的识别和分类.Basing on the echo signal to identify the underground targets is the key problem of me application of ground-penetrating radar. Data-time axial is used to reject and restrain the direct wave, and widehand correlation processing is used to filter to improve the signal to noise ratio （SNR） and extract the typical road data. The Welch power spectrum result of the typical road data is used as the input of radial basis function （RBF） network after the normalization processing, accomplishing the underground target automatic material identification and classification. Besides, it gives an effective analysis on the influence of the different cutting points to the results of target material recognition. The processing results of real data indicate that it can distinguish iron, aluminium and soil effectively.

关 键 词：径向基函数神经网络 典型道数据提取 材质识别 探地雷达 

分 类 号：TN957.52[电子电信—信号与信息处理]