In order to solve the current dynamic OD estimation problems on the background of gradual application of automated vehicle identification facilities, the relationship between dynamic OD estimation and traffic parameters under AVI environment is analyzed. The dynamic OD estimation model basing on Kalman filter algorithm is established. The coefficient matrixes of state equation and observation equation are calibrated dynamically by neural network respectively. The simulation results show that the model has higher estimation accuracy for OD pairs with great flows. The model can be adopted as one of the theoretical models for dynamic OD estimation supporting traffic control and management.