An adaptive threshold fast DBSCAN algorithm with preserved trajectory feature points for vessel trajectory clustering

Vessel navigation pattern recognition plays an important role in the research of intelligent transportation on water. Clustering using the data stored in The Automatic Identification System (AIS) is a current research hotspot. However, there are three problems in the past clustering analysis. First, the traditional Douglas-Peucker (DP) Compression Algorithm exists feature point loss and trajectory distortion when compressing trajectories. Second, Dynamic Time Warping (DTW) and the density-based spatial clustering of applications with noise (DBSCAN) algorithm require high time cost. Finally, most of the studies ignore the interaction between parameters when choosing the parameters of DBSCAN. These problems seriously affect the efficiency and accuracy of clustering. To solve these problems, this paper improves the existing methods by (1) Adaptive selection of compression thresholds and trajectory feature points for each trajectory when using the DP algorithm ensures the realism of the compressed trajectory; (2) using the Fast-DTW algorithm to improve the computation speed and ensure the accuracy of trajectory similarity; (3) Self-selection of parameter combinations based on Silhouette Coefficient (SC) scores was achieved using the similarity distribution of the trajectories in combination with an improved K-Adaptive Nearest Neighbors (KANN). The experiments show that the proposed method can greatly reduce the time cost of clustering compared to the original method and significantly outperforms the three compared algorithms in terms of clustering effect images.