Motion pattern analysis using partial trajectories for abnormal movement detection in crowded scenes

A motion pattern analysis method for abnormal movement detection is introduced. It analyses motion patterns and detects abnormal movement using partial trajectories, which can be obtained in crowded scenes and are more effective than local motion. In addition, the proposed method is able to deal with noisy data, which is a major cause of false alarms. The experimental results from real-world traffic scene datasets show that the proposed method improves on previous, local motion-based methods.