Application of regularized deconvolution technique for predicting pavement thin layer thicknesses from ground penetrating radar data

• We use the deconvolution algorithm to deal with the “thin layer problem”.
• Regularization is used to make the solution of ill-posed problem smooth.
• The maximum thickness estimation error is 4.24%.

In this paper, regularized deconvolution is utilized to analyze GPR signal collected from thin asphalt pavement overlays of various mixtures and thicknesses on a test site. By applying regularized deconvolution and the L-curve method, the overlapped interface was identified in the signal. The thickness of the thin layer was predicted with maximum error of 4.2%, which is less than 1.5 mm, a value well below the layer tolerance during construction. The study shows that the algorithm based on regularized deconvolution is a simple and effective approach for processing GPR data collected from thin pavement layers to predict their thickness.