Abstract:In order to improve the learning and analysis level of vehicle’s motion patterns, considering the characteristics of trajectories and the requirements of trajectory modeling, a modeling method for trajectories was proposed. This method consisted of two parts, which were trajectory fuzzy clustering and path modeling. Firstly, the improved Hausdorff distance was extended and used to measure the geometrical similarity between trajectories, and the improved fuzzy C-means trajectory clustering algorithm was further established to realize the clustering of vehicle trajectories. Based on the results of trajectory clustering, the path models based on discrete state were established and the Corresponding trajectory anomaly detection algorithm was proposed. Finally, the experimental results in the real scene verified the applicability and validity of the proposed method .
