To get the optimal flight routes which satisfy constraints such as mission request, navigation and security in large scope and real terrain important to enhance the unmanned aerial vehicle (UAV)'s weapon system efficiency. An improved genetic algorithm which based on direct encoding method of individual and local optimization operator is represented to solve the route planning problem. By applying the complex constraints to the evolutionary process, the algorithm can overcome the drawbacks of canonical GA. The control parameters are selected by the uniform design of the experimental method, and the result of examples has proved that the algorithm is effective.