With their high mobility and flexible configuration, unmanned aerial vehicles (UAVs), play important roles in the development of intelligent transportation in smart cities. Due to the complex three-dimensional (3D) flight environments, obtaining an optimal collision-safe flight path is a challenge with the limited computational capability of UAVs. In this paper, we propose a UAV trajectory design method for obstacle avoidance, which could be used in UAV intelligence navigation. First, we propose a cell-varying JPS method in hexagonal cells to integrate the aerodynamic constraints into trajectory designing in a two-dimensional (2D) plane. Second, we develop the cell-varying JPS with modified pruning and jumping rules and extend this modified version to trajectory-planning in the three-dimensional (3D) environments. Our simulation results demonstrate that the proposed cell-varying JPS achieves the reduction up to 4.2% in path length compared with the existing method while guaranteeing a safe flight.