The paper describes a method for obtaining information about the thickness of road pavement layers using the “Terrazond” multichannel ground penetrating radar. The method includes the use of signal pre-processing to eliminate common-mode interference on radar data of the road surface. The results of applying the methodology to data obtained during sounding of a road test site are presented. Using the example of a radar data of a test section, it is shown that three iterations of the proposed filtering contrastively highlight three layers of the road structure. The results of automated construction of the oundaries of road layers in the GeoReader program using machine learning technology are shown