Aiming at the requirements of time effectiveness and trajectory distribution of collision avoidance decision-making problem, a non-dominated sorting in genetic algorithm Ⅱ(NSGAⅡ) based on multi-objective optimization algorithm was improved. The ondemand layering strategy and the arithmetic crossover operator strategy considering the parental domination information were adopted to reduce the time complexity of the algorithm and speed up the convergence. A dynamic distributed fitness strategy was proposed to control the distribution of Pareto sets in target space. On this basis, the safety and economy objective functions of the decision-making scheme and the driver’s preference function for decision-making safety were established respectively, and the collision avoidance decision-making scheme was optimized by improving NSGAⅡ. The experimental results show that the convergence speed and distribution of the improved NSGAⅡ are improved, which proves the effectiveness and superiority of the improved algorithm. Under the four types of ship encounter scenarios constructed, the algorithm can find multiple collision avoidance decision-making schemes that take into account both safety and economy, which provide reference for the driver’s collision avoidance decision-making.