Ship Pose Estimation Based on Convex Hull: A Case Study of Ships Entering the Three Gorges Ship Lift

Accurate estimation of ship pose is important. It serves as the foundation for ship navigation decision-making. However, it is difficult to extract accurate direction information when faced with incomplete point cloud data. This study proposes a ship pose estimation method based on convex hull. By designing a geometric shape classifier, the ship point cloud cluster is categorized into symmetric or asymmetric cluster. The extracted point cloud cluster is reduced to convex hull, significantly reducing the computational burden. Then, search-based algorithms are developed for symmetric and asymmetric clusters, respectively, utilizing new criteria of minimum symmetric area difference and minimum occlusion area. The effectiveness of the method was verified on three typical ships of the Three Gorges Ship Lift. The results show that compared with principal component analysis (PCA), L-shaped fitting, and minimized occlusion region (MOR) methods, the proposed method has higher accuracy and robustness, while maintaining real-time solution speed.