An improved CBBA (consensus-based bundle algorithm) was proposed to solve the heterogeneity of task resources carried by multi-UAVs during reconnaissance and strike missions, and the requirements of tasks for heterogeneous resources. Considering the mission value, mission execution time window and range cost, a task allocation model of multi-UAV reconnaissance and attack on ground targets was established.K-medoids clustering analysis method was used to cluster multi-UAVs based on distance and the balance of carrying resources to solve the requirements of multi-UAVs for heterogeneous resource types. The attack task was generated into subtasks, and the improved CBBA was used to solve the established task allocation model. The feasibility and effectiveness of the algorithm are verified by comparison simulation experiments.