In the multi-UAV and multi-objective forest fire rescue reconnaissance scenario, addressing the issues of high complexity in task planning and the difficulty of balancing task planning speed and task execution efficiency, a mission planning method for multi-target reconnaissance is proposed. This method appropriately decouples the task planning problem to reduce complexity and divides the task planning into three parts: trajectory planning, task target allocation, and task sequence planning. A trajectory planning algorithm based on state space sampling planning and Dubins curves is proposed, which considers UAV dynamics constraints, constructs a task node connectivity graph for subsequent steps, and generates a feasible trajectory set. During the task target allocation stage, spectral space transformation is conducted. Based on the nonlinear characteristics of target group division, spectral clustering is used for rapid target allocation. A hybrid optimization algorithm combining the branch and bound method and simulated annealing algorithm is proposed to achieve task sequence planning, and the feasible trajectory set is used to generate the task execution trajectory. This method designs a decoupling algorithm for the task planning problem, enabling fast planning of optimal results. Simulation comparison experiments with other task planning methods show that this method results in shorter task planning trajectory lengths, less planning time, and task time, while the trajectories satisfy dynamics and no-fly zone constraints.