A Clustering-based Motion Understanding Method for Traffic Vehicle

Motion understanding is the classification process for time-varying data, and vehicle motion understanding under road traffic scene is a systematic research work. This paper proposed a clustering-based vehicle motion understanding method, which preprocesses the obtained motion trajectory, then uses Fuzzy C Mean Clustering to cluster the preprocessed trajectory collection and uses Hausdorff Distance to classify vehicle trajectory to be tested. On the basis of correct classification, combined with the context information of vehicle object, vehicle behavior understanding is achieved. To verify the effectiveness of our method, we have experiments with violation-oriented traffic incident behavior. The experiment results show that our method has good feasibility and robustness.