In the event that the L3 autonomous driving system fails or has difficulty handling complex traffic environments, the driver is required to takeover in an emergency, which can easily lead to traffic accidents. In order to assess the takeover risk of L3 autonomous vehicles, a takeover scenario on the urban expressway was designed and driving simulation experiments were carried out. Based on the theory of driving risk field, dynamic and static risk distribution functions were used to reflect the influence of other traffic units on the takeover risk of ego vehicle. And then, vehicle performance probability factor was introduced to indicate the probability of potential traffic accidents caused by abnormal vehicle trajectories during the takeover process, as well as considering the influence of takeover response time, a risk evaluation model of autonomous driving takeover was constructed. The model parameters were calibrated on the basis of the takeover reaction time and trajectory data obtained from the driving simulation experiments and compared with the inverse time-to-collision to verify the model. The results showed that the values of takeover risk index calculated by the model from 1 s to 9 s after the driver took over were consistent with the inverse time-to-collision. However, the root mean square error of risk index during the takeover (0.059) decreased by 37% compared to the root mean square error of the inverse time-to-collision (0.093). In summary, the constructed model can effectively assess the risk of driver takeover, and the model is more accurate than the inverse time-to-collision in describing the risk.