In order to solve the decision-making problem of UAV ground attack, the index factors affecting the threat degree of ground targets were analyzed and quantified, and a target threat assessment model based on fuzzy neural network was constructed. The target threat value and strategic benefit value are calculated respectively by using fuzzy neural network and combat constraint relationship, and the attack decision is completed by dynamically assigning weight factors of target threat value and strategic benefit value. The researched air-to-ground decision-making algorithm has the characteristics of high model accuracy, few samples and strong generalization ability, which is conducive to the fast and accurate decision-making of air-to-ground attacks. Finally, the correctness of the algorithm is verified by a simulation example. The results show that the fuzzy neural network is fast and accurate in the process of calculating the target threat value.