In order to solve the problem of that it is difficult to make reasonable use of resources in a hypersonic vehicles early warning system, an early warning resource allocation method based on bi-level programming is proposed. Firstly, the Markov state transition model of target motion state is established, and the method of threat assessment is proposed. Secondly, based on hidden Markov model and Kalman filter, a hypersonic vehicle early warning resource allocation model based on bi-level programming is proposed. The lower level planning takes the information gain under the unit resource loss as the objective function, and the upper level planning takes the risk reduction as the objective function. In order to solve the model, an improved generalized particle swarm optimization algorithm is proposed. Finally, the effectiveness and feasibility of the resource allocation method and algorithm proposed are verified by simulation analysis.