Cracks in road pavements are usual damages because of heavy and dense traffic. Cracks usually are hidden to visual inspections in the case of rigid pavements. In these cases, cracked cement concrete layers are located below the asphalt layer. Therefore, detection becomes a difficult task, being only possible in the case of major damage or even failure. As consequence, rehabilitation is complex due to the technical difficulty in finding those cracks.Non-destructive surveys are in many cases applied to detect those damages. However, the difficulty in data interpretation limits their use forcing the combination of different technologies and core samples. One of the most useful method, commonly applied to define the thickness of layers, is Ground Penetrating Radar (GPR). In order to understand better the GPR images obtained as consequence of damage in the case of rigid pavement, several tests were designed to determine the ability of GPR to non-destructively detect and quantify cracks in concrete samples covered by an asphalt layer. The laboratory experiment consisted of three testing configurations: (1) one concrete sample without cracks; (2) a model of a rigid pavement with the asphalt layer without cracks; (3) the same model with different crack widths. The test was performed with a 1.6 GHz centre frequency GPR antenna in two relative positions with respect the cracks, being always the radar line perpendicular to the crack axis. These measurements allowed also evaluate the effect of orientation of the crack with respect the antenna dipoles in the GPR images. In addition, numerical models were compared to the laboratory results. The results showed that the crack width was detected in most cases, even the thinnest ones (2 mm), being the higher amplitude anomaly corresponding to the case of the data acquired with the antenna parallel to the radar line.