For the requirement of Target type recognition in air defense combat process, this paper points out the limitations of traditional and generalized Grey Relation (GR) models in data processing, and then an improved GR model based on entropy-weight method to recognize the unknown target type is established. Firstly, in order to make full use of air targets' data exactly, the method of entropy-weight is introduced in this model to determine the index weight. Secondly, the Certainty Factor is used instead of the absolute value to more accurately describe data errors between different serial numbers. Finally, the discretization of the output of this model is made, which enhances the ability to distinguish different target types effectively. Furthermore, examples are taken to validate this model, the results show that the model is accurate, simple, high in recognition rate and of great value in application.