Abstrait

Human Crowd Behavior Analysis Based On Graph Modeling and Matching In Synoptic Video

B.Yogameena,K.Sindhu Priya

Huge video dataset captured by using various resources like surveillance cameras, webcams.etc. It is very time consuming to watch whole video manually. In this paper, Video synopsis is used to represent a short video while preserving the essential activities for a long video. In the existing methodology, usually a single moving object is splitted into a few small pieces in a continuous activity. For that Gaussian mixture model is used to detect compact foreground against their shadows. But in high-density crowds background subtraction fails due to occlusion. After detecting the occlusion, background is subtracted. Tracking is done by Centroid of the silhouette. For video synopsis more fluent tubes are generated by concatenation. Then each isolated region is considered as a vertex and a human crowd is thus modeled by a graph. After modeling the graph by using Delaunay triangulation and Convex-Hull, graph matching algorithm is developed to detect the problem of behavior analysis of human crowds. Experimental results obtained by using extensive dataset show that the proposed algorithm is effective in detecting occlusion and anomalous events using video synopsis.