Detection and localization of abnormal activities in video surveillance system

Problem of detection and localization of abnormal behaviors in crowded scenes is considered in this paper. To extract various activities of crowd, spatiotemporal Laplacian eigenmap method is proposed. Here, spatiotemporal variations of local motions are learned in an embedded space. Regular behavior...

Full description

Saved in:
Bibliographic Details
Main Authors: Momin, B. F., Fanase, V. M.
Format: Conference Proceeding
Language:English
Subjects:
abnormality detection
Computational modeling
crowd analysis
Cybernetics
Hidden Markov models
Laplace equations
manifold embedding
Speech
Speech processing
Video surveillance
Visualization
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Problem of detection and localization of abnormal behaviors in crowded scenes is considered in this paper. To extract various activities of crowd, spatiotemporal Laplacian eigenmap method is proposed. Here, spatiotemporal variations of local motions are learned in an embedded space. Regular behavior of a crowd is modeled using representatives of various activities. To detect abnormal behavior in local as well as global contexts and to localize regions showing abnormality, model of regular crowd behavior is used. Computational simplicity is the feature of this method.
DOI:10.1109/ICCSP.2015.7322886