Target Detection of Ground Penetrating Radar Actual Road Anomaly Data

Deep learning algorithm such as Y 010v3 is widely used for target detection in images, and its extension to ground-penetrating radar (GPR) road anomaly interpretation can effectively overcome the shortcomings of the traditional manual methods. However, the need for confidentiality, information islands and other types of practical problems seriously restrict the collection of real samples. In order to improve the GPR road anomaly interpretation rate, this paper uses the traditional method and the self-supervised conditional generative adversarial network (SS-CGAN) method to augment the already interpreted GPR real measurement samples. The problem of insufficient data is solved, and automatic interpretation of GPR measured road anomaly data can be achieved.