Aiming at the noise， cost and safety problems of current unmanned aerial vehicle （UAV） operation within cities， this paper divides the airspace into layers by altitudes and proposes an operation cost and crash risk model based on noise protection zones under aerodynamic and other constraints. This model satisfies the standard operation conditions of UAVs， and reduces the noise impact on the environment and the ground population， as well as operation cost and crash risk. Thus， based on the noise protection zone， the improved Dubins path planning method is used. It combines the Dubins path planning idea with the tangent line of the geometric circle， and adds node processing to enrich the optional paths for UAV operation and optimize the paths. As the Dijkstra algorithm is used as the the best path searching algorithm， and the lowest total cost is set as the goal， a UAV optimal operation path is searched and compared with the A* algorithm. Simulation experiments verify the effectiveness of the proposed model and the improved method. They reduce the noise impact of UAV operation and the operation cost， and improve the safety and efficiency of the operation. The results indicate that the optimal operation altitude of the example UAV， a biplane medium-sized UAV with a mass of 15 kg and a paddle disk area of about 1.313 m²， is 40 m， and its operation minimum total cost is 5.42； compared with the results in the other altitude layers， the total operation cost is reduced by 37.56% at most and 5.91% at least.