Vessel Re-Identification (Re-ID) has become an increasingly critical task for maritime security due to the rapid growth in the number of ships and the current lack of effective identification methods. Specifically, vessel Re-ID faces significant challenges in data collection and often encounters significant inter-class similarity, which complicates the classification and Re-ID processes. To address these challenges, we construct a large-scale vessel Re-ID dataset and propose a novel vessel Re-ID method based on part-whole hierarchies, which addresses the issue of local feature misalignment and reduces computational cost. Additionally, we introduce a comprehensive Re-ID computation framework tailored to meet the demands of industrial applications, ensuring highly accurate identity determination for the query target. Our proposed method demonstrates superior performance on the Rank@n metric, significantly outperforming current state-of-the-art techniques. These results highlight the method’s great potential for real-world applications, offering both efficiency and high precision in vessel Re-ID tasks.