The integration of sensing and communication (ISAC) using millimeter-wave (mmWave) attracted significant attentions recently in the research on Internet of Vehicles (IoV). While the ISAC technique offers advantages with increased bandwidth and improved time-frequency resolution, it encounters challenges in achieving accurate beam alignments, especially in highly dynamic scenarios. To enhance the accuracy of narrow beam alignment while mitigating the overhead in vehicular networks, this paper proposes an improved predictive beam tracking scheme based on the transceiver division structure and the bistatic radar theory. Specifically, the proposed scheme utilizes the ISAC echo signals to measure the vehicles' states and adopts an unscented Kalman filter (UKF) to reduce the overhead while ensuring the quality of service (QoS) in complicated road conditions. Simulation results show that the proposed scheme improves the prediction accuracy of angle of departure (AoD), which consequently ensures higher achievable communication rates in the IoV context.