A New System for Event Detection from Video Surveillance Sequences - Advanced Concepts for Intelligent Vision Systems - page 109

A New System for Event Detection from Video
Surveillance Sequences

Ali Wali, Najib Ben Aoun, Hichem Karray, Chokri Ben Amar,
andAdelM.Alimi

REGIM: REsearch Group on Intelligent Machines, University of Sfax, National
School of Engineers (ENIS), BP 1173, Sfax, 3038, Tunisia
{ ali.wali,najib.benaoun,hichem.karray,chokri.benamar,adel.alimi } @ieee.org

Abstract. In this paper, we present an overview of a hybrid approach
for event detection from video surveillance sequences that has been de-
veloped within the REGIMVid project. This system can be used to in-
dex and search the video sequence by the visual content. The platform
provides moving object segmentation and tracking, High-level feature
extraction and video event detection.We describe the architecture of the
system as well as providing an overview of the descriptors supported to
date. We then demonstrate the usefulness of the toolbox in the context
of feature extraction, events learning and detection in large collection of
video surveillance dataset.

1 Introduction

Image and video indexing and retrieval continue to be an extremely active area
within the broader multimedia research community [3,17]. Interest is motivated
by the very real requirement for ecient techniques for indexing large archives
of audiovisual content in ways that facilitate subsequent usercentric accessing.
Such a requirement is a by-product of the decreasing cost of storage and the now
ubiquitous nature of capture devices. The result of which is that content repos-
itories, either in the commercial domain (e.g. broadcasters or content providers
repositories) or the personal archives are growing in number and size at virtu-
ally exponential rates. It is generally acknowledged that providing truly ecient
usercentric access to large content archives requires indexing of the content in
terms of the real world semantics of what it represents.
Furthermore, it is acknowledged that real progress in addressing this challeng-
ing task requires key advances in many complementary research areas such as;
scalable coding of both audiovisual content and its metadata, database technol-
ogy and user interface design. The REGIMVid project integrates many of these
issues (fig.1). A key effort within the project is to link audio-visual analysis
with concept reasoning in order to extract semantic information. In this con-
text, high-level pre-processing is necessary in order to extract descriptors that
can be subsequently linked to the concept and used in the reasoning process.
In addition to concept-based reasoning, the project has other research activities
that require high-level feature extraction (e.g. semantic summary of metadata