Distribution Based Crowd Abnormality Detection

Abstract

The complications of abnormal behavior and behavior identification are very eminent problems in the video processing. Abnormal behavior detector can be designed by choosing the region of interest through feature detector and by tracking them over the short time period. Therefore, the detector shows the trade-off among the object tracking and optical flow. Since, various regions normally display the various types of motion pattern, we introduce Distribution Based Crowd Abnormality Detection (DCAD) which catches the statistics of object trajectories which are passing via the Spatio-temporal cube. This technique directly provides the distribution to define the frame. Also clustering is not required to build the dictionary. Besides, we exploited the motion trajectories to calculate the “power potentials” in the pixel space which defines the amount of interaction among the people. Furthermore, utilize the standard method for classification by considering SVMs (Support Vector Machines) discriminative learning method to recognize the abnormalities.