The detection tasks of the unmanned aerial vehicle (UAV) cluster for multi-objectives need to seek an optimal balance in terms of task planning speed, task planning complexity and task execution efficiency. In response to the above problems, this paper proposes a task planning method that adapts to multi-objective detection for UAV clusters. This method uses the spectral space conversion method to use the Gaussian kernel function to obtain a similarity matrix in the target allocation, which can describe the similarity between nodes and then convert it into the solution to the minimum cutting problem; in the stage of generating the task sequence, Branch and Bound (B&B) algorithm is used to generate a task sequence, and then determine the task execution order and obtain the task track. This method uses graph theory to optimize the relevant algorithms and can plan better results in a short time. Comparative experiments were conducted with other planning methods, and the results were concluded that under certain conditions, the method has certain advantages in task track length, planning time, etc.