Combining with the development of current technology, this paper proposes an improved greedy dynamic programming algorithm based on the fusion of greedy algorithm and dynamic programming algorithm for the problems of large number of station observation tasks, difficult problem modeling, and high solution complexity. First, this algorithm divides the scheduling problem into several subproblems, iteratively solves them with the objective function of maximizing the observation income according to constraints and generates an approximate optimal observation plan. Experimental results show that this algorithm has certain feasibility and practicability in solving observation task scheduling problems, and lays a solid foundation for the next step of establishing an automated station operation system.