Typical trajectory extraction method for ships based on AIS data and trajectory clustering

Abstract

Ship trajectory data is an important spatio-temporal data, which contains important semantic information and behavior pattern information. Typical ship trajectory is a representative trajectory model in ship shipping process, which describes similar behavior patterns of ship groups and is of great significance in characterizing massive trajectory features, analyzing group motion behavior and detecting abnormal ship behavior. In this paper, we propose a typical trajectory extraction method based on improved DBSCAN clustering algorithm, firstly, in the aspect of AIS data preprocessing, we propose a multi-feature dimensional trajectory segmentation algorithm to segment the ship trajectory to avoid ignoring the local important information because of studying the ship trajectory as a whole, and propose an improved DP trajectory compression algorithm with multi-feature point extraction to compress the ship trajectory to improve the data In terms of trajectory segment similarity, the Hausdorff distance algorithm is adopted and improved. In terms of trajectory clustering, the improved adaptive parameter fast DBSCAN clustering algorithm is used for trajectory clustering, and finally, the typical motion trajectory of the ship is obtained by the center-of-mass vector extraction method. Experiments with AIS data collected in the inland river basin of the Yangtze River in Wuhan show that this paper can extract typical motion trajectories in large-scale trajectory data, and improve the accuracy and efficiency of the algorithm to a certain extent, laying the foundation for further ship abnormal behavior detection.