Reliability of signal processing technique for pavement damages detection and classification using ground penetrating radar

Abstract

Ground penetrating radar (GPR) signal processing is a nondestructive technique, currently performed by many agencies involved in road management and particularly promising for soil characteristics interpretation. The focus of this paper is to assess the reliability of an optimal signal processing algorithm for pavement inspection. Preliminary detection and subsequent classification of pavement damages, based on an automatic GPR analysis, have been performed and experimentally validated. A threshold analysis of the error is carried out to detect possible damages and check if they can be predicted, while a second threshold analysis determines the nature of the damage. An optimum detection procedure is performed. It implements the classical Neyman-Pearson radar test. All the settings needed by the procedure have been estimated from training sets of experimental measures. The overall performance has been evaluated by looking at the usual receiver's operating characteristic. The results show that a reasonable performance has been achieved by exploiting the spatial correlation properties of the received signal, obtained from an appropriate analysis of GPR images. The proposed system shows that automatic evaluation of subgrade soil characteristics by GPR-based signal analysis and processing can be considered reliable in a number of experimental cases.