A Bayesian framework-based ship intention inference model is proposed in this paper for the ship intention recognition problem in maritime navigation safety. The predicted trajectories are combined with an artificial potential field model, considering the influence of COLREGS rules, to construct an algorithm for intention recognition and autonomous collision avoidance. By establishing collision avoidance waypoints and a Probabilistic Roadmap model, and integrating Bayesian intention inference and trajectory prediction models, this paper can infer the avoidance intention of the target ship by observing the current encounter situation without communication, thereby predicting the position of the other ship and determining the decision of this ship. Additionally, this paper improves the artificial potential field model, constructing a new field to ensure making evasive decisions within a sufficiently safe range. Simulation results show that the model can effectively improve the efficiency of collision avoidance algorithms, which is of great significance for preventing maritime collision accidents.