Global Abnormal Event Detection in Video via Motion Information Entropy

Abnormal event detection is a rapid development task in video analysis, which is aimed to distinguish abnormal and normal events in surveillance videos. As the normal and abnormal events have some similarities, more discriminating methods or motion information need to be explored. In this paper, to detect escaping people in different scenes, a global abnormal event detection method is proposed based on motion information entropy. To be specific, the proposed method utilizes the directions and magnitudes of motion vectors extracted from optical flow field to calculate the motion information entropy which represents the uncertainty of motion information. The normal samples are distributed around the center of a Gaussian distribution while the abnormal ones are distributed on the side. Experimental result shows that the proposed method has better detection performance than most of other classic methods with much lower computational complexity.