With the development of technology and the change of war patterns, multi-UAV systems have been widely used in battlefield. Task allocation is the key to affecting the effectiveness of multi-UAV systems. This paper establishes a multi-UAV cooperative reconnaissance task allocation problem model (MCRTAP) that is based on the background of cooperative reconnaissance of multiple targets by multi-UAV. The MCRTAP model is aimed at minimizing the reconnaissance completion time and the average time of flight, in which maximum range and cruising speed of different UAVs as well as the size of different target areas are taken into consideration simultaneously. In order to solve the MCRTAP model more efficiently, this paper proposes a particle swarm optimization algorithm based on experience pool named EPPSO. EPPSO firstly adopts the constraint-based particle swarm initialization strategy to ensure that each particle is located in the effective solution space at the initial stage. In order to let the particles explore more effective solution space, a particle position reconstruction strategy based on experience pool is presented, which uses the experience of good particles to reconstruct the position of particles that dissatisfy constraints in the evolution process. In addition, a time-varying parameter adjustment strategy is introduced to further improve the population diversity and convergence ability. The experimental results show that, compared with some state-of-the-art PSO-based improved algorithms, EPPSO can get the lowest fitness value under various experimental conditions.