The problem of cooperative task allocation for multi-UAVs in rescue scenarios is studied. Considering the different types of assistance required by survivors, a more practical combinatorial optimization model is established, and an adaptive genetic learning particle swarm optimization(AGLPSO) algorithm is proposed for this model. Firstly, according to the rescue relationship between UAVs and survivors, a real vector coding mechanism is adopted to deal with the constraints of decision variables to simplify the solution of the model. Then, the search ability of the algorithm is improved through the two cascading layers. In the first layer, the genetic learning strategy is used to generate elite particles with high quality, and the evolutionary stagnation particles are updated by the elite learning strategy to jump out of local optimum. In the second layer, the search direction of the population is guided by the elite particles, and according to the evolution speed of particle swarm and aggregation degree of particles, the adaptive evolution strategy is used to improve the searching ability of the algorithm in different evolutionary periods. The simulation results show that the proposed AGLPSO algorithm can quickly and effectively find a reasonable rescue allocation scheme.