Abstract: Millimeter-wave road monitoring radar has the advantages of all-day, all-weather, and long-distance monitoring, and plays an important role in traffic safety, traffic flow, road monitoring and other fields. However, the multipath ghost target has always affected the performance of this radar. Firstly, the echo signal model of radar was established on the basis of analyzing the propagation paths of multipath signals in highway scenarios. Secondly, the constant false alarm rate detection technique was employed to extract the relevant features of the target in the presence of multipath detection points. Additionally, the distance and angle between multipath ghost targets and road guardrails, the continuity of detection point clouds and other features were analyzed. Finally, a neural network-based ghosting suppression method was proposed based on these identified features. Through empirical data analysis, the accuracy of target detection and tracking by millimeter-wave road monitoring radar was verified. The results indicate that the utilization of the proposed 8 features effectively mitigate the presence of multipath ghosts, with a correct recognition rate of 95.74% and an overall recognition accuracy rate of 90.70%. The method enables effective monitoring and precise positioning of vehicles.