Development of regularization methods on simulated ground-penetrating radar signals to predict thin asphalt overlay thickness

Abstract

The range resolution of ground-penetrating radar (GPR) signal is important in thin asphalt overlay thickness estimation. In this paper, regularized deconvolution is utilized to analyze simulated GPR signals to increase their range resolution. Four types of regularization methods, including Tikhonov regularization and total variation, were applied on noisy GPR signals; and their performance was evaluated in terms of accuracy in estimating distance of close impulses. The L-curve method was used to choose the appropriate regularization parameter. The total variation regularization method and zeroth-order Tikhonov regularization outperform first-order and second-order Tikhonov regularization in terms of average asphalt layer thickness estimation error and the standard deviation of the error. An example of the field GPR data is provided to validate the proposed algorithm. The study shows that the algorithm based on regularization is a simple and effective approach to increase the GPR signal range resolution with presence of noise in the case of thin asphalt overlay thickness prediction.