Abstract: Considering aerial reconnaissance operations with multiple starting points and multiple destination points, on the basis of task requirements, plane performance and constraints, the model for reconnaissance scheduling and path planning is built from perspective of the multiple travelling salesman problem. By improving the genetic algorithm, a better solution is obtained to achieve the optimal performance. To measure the importance of airports, the loss of performance is computed by means of node removal. This method combines importance analysis with task planning, which provides a useful reference for the research of targets ranking driven by tasks.