This paper advances a method to use support vector machine(SVM) to process and analyze the roadbed detection data obtained by ground penetrating radar(GPR),so as to achieve the quantity interpretation of roadbed disease recognition.Some actual roadbed detection data obtained by GPR was processed and analyzed by this method,and the results indicated that the nicety degree of roadbed disease recognition was up to above 90%,proved that the SVM method can result in good effect and high accurate degree in roadbed detection.