TrajCleaner: an end-to-end framework for systematic cleaning and completing vessel trajectory data

Abstract

Reliable vessel trajectory data are fundamental for maritime situational awareness and intelligent port management, yet Automatic Identification System (AIS) records frequently suffer from drift errors, temporal gaps, and fragmented sequences. Existing studies on AIS trajectory cleaning and gap completion remain limited by fragmented workflows, weak integration between detection and completion, and insufficient adaptivity across heterogeneous maritime environments. To address these challenges, this paper proposes TrajCleaner, a systematic end-to-end framework for trajectory anomaly detection, repair, and event-based completion. This framework integrates adaptive multi-feature detection to identify missing or abnormal segments across diverse vessel types and port operational regimes. A multi-factor weighted similarity mechanism enables robust candidate retrieval, while a spatio-temporal joint completion strategy enforces port-specific navigational constraints and performs density-aware temporal allocation to ensure behavioral plausibility. Unlike prior approaches that treat drift correction, gap detection, and interpolation as independent tasks, TrajCleaner provides a unified pipeline from anomaly recognition to event-level trajectory completion. Experiments using AIS data from the port of Los Angeles (LA) demonstrate that TrajCleaner substantially enhances detection accuracy, completion fidelity, and temporal realism compared with state-of-the-art baselines. By generating eventized and behaviorally coherent vessel trajectories, TrajCleaner improves the analytical depth of downstream maritime applications, including vessel behavior profiling, port operational efficiency assessment, and channel and berth planning. These capabilities provide a more reliable data foundation for predictive, adaptive, and decision-support systems in next-generation smart ports, ultimately strengthening maritime safety, operational efficiency, and environmental sustainability.

More information

Identifiers

ORCID for Huanhuan Li: orcid.org/0000-0002-4293-4763

Catalogue record

Export record

Altmetrics