Paper Info
Title
A macro-observation scheme for abnormal event detection in daily-life video sequences
Abstract
We propose a macro-observation scheme for abnormal event detection in daily life. The proposed macro-observation representation records the time-space energy of motions of all moving objects in a scene without segmenting individual object parts. The energy history of each pixel in the scene is instantly updated with exponential weights without explicitly specifying the duration of each activity. Since possible activities in daily life are numerous and distinct from each other and not all abnormal events can be foreseen, images from a video sequence that spans sufficient repetition of normal day-to-day activities are first randomly sampled. A constrained clustering model is proposed to partition the sampled images into groups. The new observed event that has distinct distance from any of the cluster centroids is then classified as an anomaly. The proposed method has been evaluated in daily work of a laboratory and BEHAVE benchmark dataset. The experimental results reveal that it can well detect abnormal events such as burglary and fighting as long as they last for a sufficient duration of time. The proposed method can be used as a support system for the scene that requires full time monitoring personnel.
Year
DOI
Venue
2010
10.1155/2010/525026
EURASIP J. Adv. Sig. Proc.
Keywords
Field
DocType
proposed macro-observation representation record,distinct distance,abnormal event detection,daily-life video sequence,daily work,full time monitoring personnel,energy history,abnormal event,daily life,macro-observation scheme
Computer vision,Support system,Computer science,Constrained clustering,Artificial intelligence,Pixel,Macro,Machine learning,Centroid
Journal
Volume
Issue
ISSN
2010,
1
1687-6180
Citations 
PageRank 
References 
2
0.39
47
Authors
2
Name
Order
Citations
PageRank
Wei-Yao Chiu1416.23
Du-Ming Tsai297068.17