基于经验模态分解的低信噪比探地雷达数据处理

阐述经验模态分解原理及复信号分析理论,着重讨论EMD分解应满足的条件及具体分解过程.应用EMD方法对单道GPR数据及GPR正演加噪剖面分别进行分解,得到从高至低不同频率范围的本征模态函数GPR图.然后,以湖南长沙黑麋峰抽水蓄能电站进厂交通隧洞实测GPR剖面为例,首先对该剖面进行EMD分解去除部分噪声,再利用Hilbert变换求取GPR剖面复信号,并提取瞬时振幅、瞬时相位、瞬时频率3个参数,绘制出相互独立的瞬时参数剖面图.研究结果表明:EMD分解对于低信噪比GPR数据具有较强的适应性,把EMD分解应用于含噪的雷达信号,并结合GPR复信号分析技术得到的“三瞬”信息,避免了使用单一时距剖面分析所造成的解释偏差,可以较好地实现对低信噪比GPR数据的噪声去除,突出雷达剖面中异常体特征,达到提高GPR信号分析效果及解释精度的目的.

GPR signal processing under low SNR based on empirical mode decomposition

The theory of empirical mode decomposition and complex signal analysis technology were described.The conditions of EMD and its process were disscussed.The EMD method was applied to decompose the single trace data and GPR forward profile with addition noise.It could obtain the intrinsic mode function pictures with different frequencies from high to low.Then,the GPR profile could be acquired in the transportation tunnel which was open into the plant of Heimi peak pumped storage power station.Firstly,the profile was decomposed with the EMD method to eliminate partial noises.Secondly,the Hilbert-Huang was transformed to calculate the complex signal of GPR profile,and three parameters such as the instantaneous oscillation amplitude,the instantaneous phase and the instantaneous frequency were extracted.Thirdly,the mutually independent instantaneous parameters profile pictures could be drawn up.The results show that empirical mode decomposition has stronger adaption to the low signal to noise ratio(SNR).EMD is applied into the radar signal with noise,and combined with the GPR three instantaneous parameters,it can avoid the explanation error which is caused by using single time distance profile analysis,and elimilate the noise of GPR data under low SNR,which can stress the unusual information and improve the GPR singal analysis effection and precision.