Evaluating the autonomous capability of groundC-attack unmanned aerial vehicle (UAV) under the condition of combat mission is one of the key problems to be solved urgently in the combat use of UAV. According to the operational process and characteristics of groundC-attack UAV, five autonomous capability influencing factors closely fitting its operational characteristics are selected, including perceptual detection, planning and decisionC-making, operational execution, security management and learning evolution, to build an evaluation index system of autonomous capability of groundC-attack UAV for the whole mission process. An autonomous capability evaluation model is established based on Bayesian network, the prior probability of root node is determined by improved entropy weight method, and an example is simulated by Bayesian software Netica. Aiming at the autonomous capability of groundC-attack UAV before, during and after mission, three reasoning modes of causal reasoning, truncation analysis reasoning and influencing factor reasoning are used for simulation verification and reasoning analysis. According to the simulation verification results, the dynamic adjustment suggestions of autonomous capability in each stage are given.