Dispersion Compensation Strategy Based on Sparse Bayesian Learning in Terahertz Nondestructive Evaluation

Terahertz (THz) technique, as a potential nondestructive evaluation (NDE), has emerged great potentials for the high resolution characterization of various non-metallic materials due to its superior detection accuracy and high sensitivity. However, when THz wave penetrates the material, the attenuation and dispersion will inevitably appear, especially for the thickness and high loss materials, which will degrade the minimum resolvable performance of THz wave and limit its wide application in super resolution characterization. Therefore, in this work, a novel method based on the sparse Bayesian learning is proposed to address the dispersion problem in THz NDE. First, the THz sparse dispersion model is established based on the prior knowledge of THz echo signal. Second, the double Gaussian mixture model (DGMM) is used to establish a parametric dispersion dictionary and a parametric non-dispersion dictionary. Then, the sparse Bayesian learning (SBL) method is applied to solve the sparse inverse problem. Finally, numerical simulations and experiments are performed to validate the applicability and effectiveness of the proposed strategy for suppressing the dispersion of THz wave in THz NDE.