Channel Extraction of Ship Trajectories Based on Improved DBSCAN

The vast expanse of the seas around Hainan island generates massive amounts of AIS data every day. Therefore, clustering trajectory data to extract ship navigation patterns is a challenging problem. In this paper, real AIS data collected by satellite from the seas around Hainan Island in November 2023 is selected. During the data pre-processing stage, trajectory extraction, trajectory interpolation and trajectory compression are performed on the selected AIS data. Additionally, in order to calculate the Dynamic Time Warping (DTW) distances more accurately, we add speed and course information other than the longitude and latitude, upon which a distance matrix is constructed. Meanwhile, the k-distance curve is utilized to select the adaptive eps to obtain better results. This paper selects the silhouette coefficient as the evaluation index, the experimental results indicate that the algorithm selected in this paper has better results compared to other methods.