Sparse Representation-Based Video Anomaly Detection Approaches

Depending on the specific scene in consideration, the abnormal event detection can be classified into those in crowded scenes and those in uncrowded scenes. Generally, crowd abnormal behaviors cannot be treated as a simple collection of individual behaviors in the form of trajectories due to the occlusion that happens among them. One solution would be to treat the crowd as a single entirety in a specific scene and detect the anomaly by analyzing the motion of the crowd in the sense of the dynamics emanating from the entire crowd. However, more often than not, it can happen that, in situations where the motion of a crowd is random and the crowd motion pattern is unstructured, the solutions proposed for structured crowded abnormal behavior detection may not work effectively enough. Further, currently, the developments achieved in the field of human behavior modeling and understanding of a crowd remain immature. Therefore, different from most existing anomaly detection methods, in this chapter, we introduce sparse representation based video anomaly detection approaches which are techniques developed for crowd abnormal behavior detection on its own.