独立分量分析在探地雷达信号处理中的应用初探

A primary study of independent component analysis for GPR signal processing

针对探地雷达(Ground Penetrating Radar,GPR)信号信噪比低、背景杂波强,事先对探测目标的信息所知甚少,近乎处于“盲”状态,因而实际处理难度大等实际问题,提出了将独立分量分析(In-dependent Component Analysis,ICA)这种盲信号处理技术应用于GPR信号处理,并利用ICA中的Fast-ICA算法,对ICA法在GPR信号处理中的应用进行了初步探索,实现了GPR信号中弱目标信号和强背景杂波的有效分离,并初步解决了对所分离目标信号的正确排序,以及由ICA方法本身带来的所分离目标独立分量信号符号的不确定性等问题,使GPR信号信噪比大幅提高,从而使GPR的目标检测性能也得以显著改善。在对诸如地雷、地下管线等局部目标的时域有限差分法(Finite-dif-ference Time-domain,FDTD)、仿真GPR数据和室外试验观测GPR数据进行ICA处理后,都取得了理想的结果。

Ground penetrating radar （GPR） signal processing has all along been a hard task because of its poor signal-to-noise ratio （SNR） of the raw GPR signals and almost ＂blind＂ about the latent targets what we want to detect beforehand. A new approach that uses Independent Component Analysis （ICA）, a newly developing blind signal processing （BSP） technique, to GPR signal processing has been proposed in this paper. With the help of FastICA algorithm, it has successfally extracted quite weak target signals from the raw GPR data with much strong background. So the SNR of the GPR signal and the detection performance of the GPR have both improved greatly. The ambiguities of rightly sequencing and the sign uncertainty problems of separated target signal components in ICA practical use have also been overcome. Two examples, FDTD simulated and field-test GPR data for local targets, such as landmines and underground pipelines, to be processed by FastICA, are included to demonstrate the effectiveness of the approach. Quite ideal results are got from them.