Abstract: Objectives Addressing the requirement for precise sensing during ship berthing and unberthing at close range, we propose a high-precision positioning method based on the integration of three-dimensional LiDAR and inertial navigation systems. Methods Initially, we develop a multi-coordinate conversion and time registration algorithm for ship navigation, achieving temporal and spatial unification of various sensing devices while correcting collected point cloud data. Subsequently, we introduce an enhanced Random Sample Consensus (RANSAC) berth shoreline detection method. This method mitigates the impact of point cloud data randomness on shoreline detection accuracy by considering the berth elevation and optimizing the clustering cost function. The identified shoreline intersection serves as a characteristic point for berthing, aiding unmanned ships in positioning. Finally, the effectiveness of the proposed method is validated through real ship experiments. Results Experimental outcomes demonstrate that this positioning method yields positioning information for ships with an error of less than 1 m. It exhibits strong adaptability to various interferences in real-world environments. Conclusions This paper proposes a berthing and unberthing positioning method, i.e., improved RANSAC for ships that can maintain high positioning accuracy even when the quality of the point cloud data is poor.