Abstract: Unmanned aerial vehicles (UAV) are gradually used in urban patrol and defense due to their advantages of low cost, high dynamics, and low deployment. To improve the efficiency of heterogeneous UAV trajectory planning, first, we establish a multi-UAV mission planning model considering UAVs' mission execution rate, trajectory cost, and impact cost is established. Second, to address the problems where the traditional optimization algorithm easily falls into the local optimal solution and poor uniformity, we introduce a differentiation strategy and Levy flight strategy into the crow search algorithm (CSA) to improve the algorithm and propose the hybrid difference crow search algorithm based on the Levy flight strategy (LDCSA). Then, we add the pruning processing and logistic chaotic mapping mechanism into the rapidly-exploring random tree (RRT) algorithm and perform track initialization using the improved RRT algorithm for track initialization. Finally, we establish a three-dimensional city model for simulation experiments and compare the proposed algorithm with particle swarm optimization, simulated annealing, and CSA algorithm. The simulation results show that the algorithm can improve global convergence and robustness, shorten the convergence time, improve UAV execution coverage, and reduce energy consumption and has advantages in solving multi-UAV trajectory planning problems.