中文摘要：

针对探地雷达数据采集和处理量大以及压缩感知算法检测异质体时稀疏字典创建困难的问题，提出基于地埋异质体回波特征的稀疏字典创建方法。根据异质体具有各向异性和因填充物不同电磁波照射下具有不同的散射特性，结合经验模态分解和小波分解提取数据特征的优点，提取回波特征分量，建立稀疏字典。采用压缩感知重构算法，实现地埋异质体的检测。实验结果表明，稀疏字典创建方法可将待处理数据量减少到原来的10％。

英文摘要：

In view of the problem that the amount of the data collection and processing of ground penetrating radar （GPR） is large and that it is difficult to create the sparse dictionary with the reflection wave when CS algorithm is used to detect the heterogeneity, a method of creating sparse dictionary is proposed. As the buried heterogeneity has typical properties of ani sotropy and is filled with different material, it will present different scattering characteristics of transmission under the irra- diation of electromagnetic. Based on this sparse characteristic of the buried heterogeneity, the method of creating the sparse dictionary is proposed, which is combining the advantages of empirical mode decomposition （EMD） and the wavelet decom- position to extract feature components of the echo of the buried target, and then the feature components will be used to cre ate the sparse dictionary. The CS reconstruction algorithm is used to detect the buried heterogeneity. The experimental re- sult shows that the method can realize the detection of buried heterogeneity with only 10 percent of the source data.