卡亨南-洛维变换法提高探地雷达信号信噪比研究

Primary Study on Ground Penetrating Radar Signal-to-Noise Ratio Enhancement via Karhunen-Loéve Transform

针对探地雷达(GPR)信号中背景杂波强、分布范围广,信号信噪比低的特点,提出一种用卡亨南-洛维变换法对GPR信号进行多道处理的方法.该方法充分利用了多道GPR数据中背景杂波、目标信号以及随机噪声等成分所具有的不同相关性这一重要的二阶统计特征,根据多道数据相关矩阵特征值的大小与数据中各成分相关性强弱的对应关系,实现了目标信号与背景杂波和随机噪声的有效分离.对时域有限差分仿真GPR数据和实测GPR数据的处理结果表明,与针对消除直达波影响的时窗法和平均相减法相比,该方法对直达波的去除更彻底(不会有拖尾效应),抗噪能力更强,信噪比改善更显著,经处理后的信号信噪比可提高30 dB以上.

In order to improve the quite poor signal-to-noise ratio （SNR） of the raw ground penetrating radar （GPR） data and increase the GPR detection performance, a new multichannel signal processing technique using Karhunen-Loéve transform to eliminate the strong GPR background and suppress the random noise in the raw GPR data simultaneously is presented. It fully uses the different second-order statistic characteristics of their correlations on the background, random noise and target signals in multichannel GPR data. The target signals are successfully separated from the background and random noise according to the corresponding relation of the eigenvalue’s magnitude of the correlation matrix and correlation strength of each component in the multichannel GPR data. The finite-difference time-domain （FDTD） simulated and practical field-test GPR data are processed to demonstrate the effectiveness of this technique. Compared with the most commonly used methods, such as time gating and ensemble average subtraction, etc. , it can more thoroughly remove the background and random noise simultaneously from the raw GPR data. Practical GPR data processing result shows that the SNR improvement factor can increase by more than 30 dB.