A Multisource Data-Driven Method for Extracting Vessel Berthing and Anchoring Regions to Support Sustainable Inland Waterway Management

Identifying abnormal vessel stay regions in inland waterways is important for navigational safety, effective supervision, and the sustainable use of limited berthing and anchorage resources. However, existing studies often rely solely on automatic identification system (AIS) data and lack the semantic information needed to distinguish different types of stay regions. This study proposes a multisource data-driven method for extracting and classifying vessel stay regions in the Three Gorges Reservoir. A parameter-adaptive iterative DBSCAN algorithm is first used to cluster vessel stay points, yielding 1006 valid stay regions. Spatial overlap analysis with mapped berths and anchorages classifies 567 regions as berths, 340 as anchorages, and 99 as abnormal stay regions. Shoreline information extracted from remote sensing imagery using a region-growing method is then used to further classify the abnormal regions into 58 abnormal berthing regions and 41 abnormal anchoring regions. Finally, prolonged stays and mismatches between berth functions and vessel types are identified across different regions, vessel types, and vessel-size groups. The proposed method enables fine-grained semantic characterization of vessel stay regions and supports vessel behavior monitoring, waterway regulation, and sustainable inland waterway management.