A ship trajectory clustering and typical trajectory extraction framework based on Grey Wolf Optimized DBSCAN

Abstract

Ship trajectory clustering based on the Automatic Identification System (AIS) is important for traffic safety and efficiency. However, current clustering algorithms have major limitations, similarity measures ignore the combined effects of multiple attributes, and trajectory compression prioritizes a high compression rate at the cost of losing essential trajectory features, which may lead to insufficient capture of navigation patterns. To address these issues, we propose a framework named Grey Wolf Optimizer with Density-Based Spatial Clustering of Applications with Noise (GWO-DBSCAN). Based on processed AIS data and the Douglas–Peucker (DP) algorithm, we design a classification logic to assign adaptive compression thresholds, thereby retaining key trajectory points. Trajectory similarity is calculated by combining Hausdorff distance with an improved Dynamic Time Warping (DTW) algorithm. The Grey Wolf Optimizer searches for the optimal weights of different distance matrices and the best clustering parameters. Finally, for each clustering result, typical trajectories are extracted and validated using a principal path extraction strategy. Experiments with real AIS data show that our method outperforms baseline models in clustering metrics such as the Silhouette Coefficient and Davies–Bouldin index, demonstrating its rationality and effectiveness in improving clustering accuracy and extracting main navigation routes.