Aiming at the problem of three-dimensional dynamic route planning for unmanned aerial vehicles(UAV), an improved constrained differential evolution(CDE) algorithm is proposed so as to meet the requirements of instantaneity and dynamic search accuracy. Firstly, this paper formulates the route planning problem for the UAV as a constrained optimization problem and constructs the objective functions and constraint functions according to the constraints of flight and threats. Then, the diversity, convergence and accuracy of the algorithm are improved through introducing generalized opposition-based learning and adaptive ranking mutation operators into the CDE algorithm. Finally, the adaptive trade-off model is applied to handle the constraints in each state, and the information of elite individual is fully utilized to achieve a reasonable conversion of the fitness. Experiment results and comparisons with 3 state-of-the-art CDEs show that the proposed method is able to plan a safe flight route which can implement static and dynamic threat avoidance effectively and realize terrain following. Compared with other three algorithms, the presented method has the advantages of terrific optimization performance, strong robustness, good convergence and high reliability.