A Fréchet Distance -Hierarchical Density-based Spatial Clustering of Applications with Noise for Ship Trajectory Clustering and Route Recognition

Nowadays, Maritime transportation remains vital component of global trade, with majority of international cargo relying on routes of sea for transportation. The main significant role in ship supervision is to record a vast amount of data related to navigation about ships, information about ships movement which is done by Automatic Identification System (AIS). The existing methods for ship trajectory clustering are affected by outliers which caused more noise in the data makes it difficult to cluster the ships trajectory and recognition. Therefore, this paper proposes Fréchet distance (FD)-Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) approach for ship trajectory clustering and route recognition which is employed on AIS dataset and effectively predicted the route. The input AIS dataset consists of static and dynamic information related to ships then the data is cleaned and repaired using cubic Hermite interpolation to remove outliers and noise from the data; then, these data are used to calculate the distance of ship trajectories with the help of Fréchet distance (FD). The HDBSCAN approach is used to cluster the ship trajectories and the prediction of route is done by using Long Short-Term Memory (LSTM). The proposed FD-HDBSCAN achieved better results in terms of accuracy (98.3%), precision (97.3%), recall (97.2%) and F1-score (96.6%) when compared with existing Extreme Gradient Boosting (XGBoost).