Trajectory Similarity Measure Design for Ship Trajectory Clustering

Abstract

Trajectory similarity measure is an important factor in Ship trajectory clustering, its design influences the number and shape of the cluster. In order to design a trajectory similarity measure appropriate for ship trajectory clustering. The author of this paper proposed an improved trajectory similarity measure based on symmetrized segment-path short (SSPD), it calculates the distance in time series and enhance the weight of endpoints, solves the problem of SSPD cannot identify trajectory direction and endpoints inconsistent. To prove that the improved trajectory similarity measure performs better than other similarity measures, the author designed two experiments. In the first experiment, two similar ship trajectories were selected, then change their sampling rate, point position, trajectory length and trajectory direction, compared with other trajectory similarity measures, observe the distance change. In the second experiment, based on real Automatic Identification System data, the improved SSPD is compared with Dynamic Time Warping on ship trajectory clustering. Experiment shows the improved SSPD can get a better clustering result.