An improved dynamic window method is proposed to solve the obstacle avoidance problem of intelligent ships in offshore waters that cannot effectively avoid when encountering unknown dynamic obstacles. Firstly, an autonomous collision avoidance decision-making model was constructed to judge whether the unmanned boat had a collision hazard, and the encounter situation was reclassified. To ensure that the USV can safely pass through the moving obstacle vessel, a collision risk assessment function considering the velocity of dynamic obstacles is introduced. Moreover, a collision avoidance rule function is incorporated into the evaluation function of the collision avoidance algorithm to ensure that the USV adheres to collision avoidance regulations. Ultimately, the simulation results confirm that the enhanced dynamic window method can safely avoid collisions while adhering to collision avoidance rules. This paper proposes an effective solution for collision avoidance behavior of intelligent vessels in near shore waters when facing unknown dynamic obstacles, providing insightful recommendations for subsequent research in maritime sectors.