Morphological component analysis based on mixed dictionary for signal denoising of ground penetrating radar

Abstract

Forward modelling is applied to simulate the ground penetrating radar (GPR) detection environment, and a modified morphological component analysis (MCA) algorithm is applied to GPR signal denoising. Finite-difference time-domain (FDTD) method is used to perform finite difference approximation to the space and time derivatives of Maxwell's equations. Under the forward simulation framework, the MCA algorithm applies a sparse dictionary to decompose the GPR signal. However, clutter is not represented as there is no corresponding sparse dictionary, the clutter is removed when the signal is reconstructed. The core of the MCA is to select a suitable dictionary. The combination of undecimated discrete wavelet transform (UDWT) dictionary and curvelet transform dictionary(CURVELET) is selected. The improved MCA algorithm is compared with singular value decomposition (SVD) and principal component analysis (PCA), to confirm the high performance of the proposed algorithm.