This paper proposed a dynamic collision avoidance method based on Artificial Potential Field (APF) algorithm for unmanned ships, and improves the defects of traditional APF respectively: In order to make unmanned ships perform collision avoidance operation safely and comply with good seamanship and the Convention on the International Regulations for Preventing Collisions at Sea, 1972 (COLREGS), this paper introduces a ship domain priority judgment encounter situation, so that the algorithm can make collision avoidance operations compatible with the actual; For the path interference and target unreachability problems, this paper introduces a new attractive potential field function for improvement, which divides the attractive potential field of the target point into multiple segments simultaneously. Meanwhile, the repulsive force on own ship is reduced when own ship is close to the target point, to avoid the path interference and overcome the target unreachability problem. Finally, for the path oscillation problem, this paper refers to partial ideas of the Dynamic Window Approach (DWA) to integrate the potential field based on the traditional APF. By comparing with the traditional APF algorithm, the overall smoothing degree is improved by 71.8% comprehensively, and the effectiveness and superiority of the algorithm are verified.