Automatic video activity recognition

Abstract

With the demands on public security and the availability of large storage systems, an increasing number of video surveillance systems are being deployed all over the world to help people detect interesting target events. However, most of these systems require intensive human monitoring, or require human operators to review video footage corresponding to extended periods of time, only to find a few short clips that are of interest. The problem fosters a demand of an automatic computer surveillance system, which can assist the human operators in identifying possible interesting events. This challenge has attracted researchers from different domains, leading to a variety of proposed approaches, particularly in the field of human activity recognition. These approaches vary in the choice of representation and methodologies as well. This chapter gives a survey and reviews the state of the art approaches to automatic human activity recognition in videos. © 2011 Springer-Verlag Berlin Heidelberg.