In the highly complex and intensely adversarial air combat environment， current unmanned aerial vehicle intention assessment methods generally face challenges such as strong subjectivity of fusion rules， ignoring the time correlation of relevant attributes， and insufficient defect information processing means. To address these issues， this paper proposes a dynamic evidence network method that integrates defect information correction and subjective and objective rules. Firstly， focusing on the strong correlation of continuous variables in the time dimension， a spatiotemporal fusion evidence network model was modularly constructed. Subsequently， by introducing the LSTM trajectory prediction technology， a defect evidence correction mechanism was established， which significantly improved the accuracy and integrity of information. Finally， objective rules were designed based on statistical calculation methods， and combined with subjective experience， a library of subjective and objective rules was constructed. Based on the above improvements， simulation experiments were conducted. The results confirm that the proposed mechanism of defect information correction， subjective and objective rules， and dynamic fusion have significant advantages in improving the accuracy and credibility of intention recognition results.