In this paper, we propose object-based predictive virtual lane (OPVL) using the flow of preceding vehicles. The proposed algorithm generates a virtual lane using the kinematic model of the ego vehicle and the location information of the preceding vehicles from radar and/or V2X. We also propose the criteria for selecting the appropriate vehicle for use in OPVL. The proposed methods were validated through actual vehicle test using three vehicles on a test road in Yeoju smart highway in Korea.