Abstract:  In order to improve the signaltonoise ratio (SNR) of ground penetrating radar (GPR) data and reduce the random diffraction energy caused by random perturbation, twodimensional variational mode decomposition (2DVMD) is introduced into the noise reduction processing of 2D GPR data. First, the GPR data is processed by 2DVMD, and the intrinsic mode function (IMF) components and their corresponding frequency and wave number spectra are analyzed to determine the type of each echo showed in the radar profile. Then, the crosscorrelation coefficients between the IMF components and the original data are calculated to determine the signal mode and noise mode, and the signal mode is reconstructed to obtain data after noise reduction. Synthetic and practical data tests demonstrate that compared with the traditional 1DVMD method, the peak SNR of the forward recording with noise after 2DVMD filtering increases from 6.44 dB to 7.72 dB. The newly developed method can get significantly improvement on the SNR of GPR data, and obtain radar profiles with better event continuity.

Key words:  , ground penetrating radar (GPR), twodimensional variational modal decomposition (2DVMD), frequency and wave number spectrum, crosscorrelation coefficient, noise reduction