Considering the actual navigation requirements of unmanned surface vehicles, the planned path should meet the requirements of smoothness and economy. Therefore, a new path planning strategy based on improved crow search algorithm combining straight line and circular arc turns is proposed. Firstly, a new path fitting method is proposed to optimize the number of turning points and handle the problem of arc transition at turning points. This method overcomes the problem of frequent direction adjustment caused by B-spline curve paths for unmanned surface vehicle, while ensuring that unmanned surface vehicle can achieve steering while maintaining a stable speed, thereby improving navigation stability and economy. Then, based on this, an improved crow search algorithm is proposed to optimize the location of path turning points. The improvement of the algorithm is mainly reflected in three aspects: using a reverse learning strategy to optimize the initial population to improve the quality and diversity of the initial population; A dynamically changing awareness probability is proposed to improve the global search ability of the initial segment and the local search ability of the final segment of the algorithm; The Levy flight strategy is used to improve the random search method to improve the directionality and effectiveness of the search. The simulation results show that the proposed new path fitting method is superior to the B-spline curve fitting method and the straight line segment fitting method; Based on this fitting method, the improved crow search algorithm, the standard crow search algorithm, the differential evolution algorithm, and the genetic algorithm are used to optimize the location of the path turning point. The iterative calculation and variance analysis results show that the proposed improved crow search algorithm has higher convergence accuracy and robustness compared to the other three algorithms, and can more effectively handle practical problems of unmanned surface vehicle path planning.