In the vehicle heterogeneous network with noise and interference, the current vertical handoff algorithms based on decision tree have the problem of low handoff accuracy. In this paper, the decision processes of current algorithms are analyzed in detail and the formulation of false decision probability is given. Firstly, the Kalman filtering algorithm is employed to obtain the more accurate network attribute values according to the predicted values, the current values, and their noise deviations. Secondly, a probability threshold interval method is proposed to do a twice detection to the situation of the attribute value which is near the threshold. Simulation results show that the proposed algorithm can improve the accuracy of handoff decision and the total network throughput, and can also reduce the ping-pong effect and the failed handoff. Meanwhile, it still keeps the same-ordered time complexity with the traditional algorithms.