Shearlet变换在GPR数据随机噪声压制中的应用

随机噪声是探地雷达(ground penetrating radar,GPR)数据处理存在的主要问题之一,直接影响到GPR数据后续处理及最终解释的准确性和可靠性。为了有效地去除随机噪声,同时更好地保留GPR信号的有效信息,本文提出基于Shearlet变换的GPR数据随机噪声去除方法。作为一种非自适应多尺度、多方向性的几何分析方法,Shearlet变换能够近乎最优地表示含奇异点的高维曲线。在Shearlet域,GPR数据能够得到更加稀疏的表示,通过阈值去噪的方法,有效地去除了随机噪声,使信噪比提高了4dB,最大程度地保留了GPR有效信号。利用理论和实际数据进行验证,体现了Shearlet变换阈值去噪方法的有效性和准确性。

Application of Shearlet Transform for Suppressing Random Noise in GPR Data

Random noise is one of the serious problems encountered during the GPR data processing.It directly influences the accuracy and reliability of the processing results.In order to remove the random noise effectively and retain the useful information,the authors propose a random noise suppression method in GPR data by using Shearlet transform.As a non-adaptive geometric-analysis method with multi-directions and multi-scales,Shearlet transform can be used to approximately represent the high-dimensional curves with singular points.In Shearlet domain,GPR data appear more sparsely.Through the threshold de-noising method,random noises can be suppressed effectively;so that the signal to noise ratio (SNR) is improved by 4 dB,and the useful information is retained to the maximum extent.The effectivity and accuracy of the Shearlet-transform threshold de-noising method are validated by the theoretical and practical data.