The multi-unmanned aerial vehicle (UAV) task allocation problem in irregular regions requires effective division of the irregular region and reasonable allocation of multiple UAVs to complete the tasks. It is a complex allocation problem with multiple constraints. In this paper, we first use the centroid Voronoi diagram to achieve uniform division of the irregular region, which improves the efficiency of multiple UAVs to complete tasks collaboratively. Secondly, to solve the problem that multi-objective particle swarm optimization (MOPSO) easily falls into local convergence, we propose a hybrid algorithm that combines heuristics and reinforcement learning, that is, the Quality-learning (Q-learning) is introduced as a means to enable the particle swarm to explore local optima more effectively, thereby enhancing the local search capability of MOPSO. It helps MOPSO obtain the global optimal solution faster and more efficiently. Finally, the experiment results show that the method we proposed has the advantages of outstanding optimization ability, and can effectively improve the mission completion efficiency of UAVs in irregular areas.