SUNAR Surveillance Network Augmented by Retrieval - Advanced Concepts for Intelligent Vision Systems - page 155

(a) (b)

Fig. 1. An example of a successful camera pair handover

The idea was to create an automated system for object visual detection, track-
ing and indexing that can reduce the burden of continuous concentration on mon-
itoring and increase the effectiveness of information reuse by a security, police,
emergency and firemen (or military) and to be useful in the accident investi-
gation. The task is to perform the analysis of the video produced by a camera
system with non-overlapping field of views. The analysis, based on cleaned, inte-
grated, indexed and stored metadata, is of two types - on-line used for identity
preservation in a wide area; and off-line to query the metadata of the camera
records when an accident, crime, a natural or human disaster (war) occurs.
In 2006, we have started to develop an IR-based multi-camera tracking sys-
tem to be at the top of the state of the art. We have taken part in several
projects (CARETAKER [4]) and evaluations (TRECVid [19]) concerning similar
problems. However, the AVSS 2009 Multi-Camera Tracking Challenge [20] was
the first evaluation campaign that used the annotated Multiple-camera Track-
ing (MCT) Dataset from the Imagery Library for Intelligent Detection Systems
(i-LIDS) provided by Home Oce Scientific Development Branch (HOSDB) of
the UK [16]. We have used the MCT video data and annotations to train and
evaluate the SUNAR performance and it performed comparably well.
The paper is organized as follows . The introduction presents our motivation
and ideas. An overview and design of the SUNAR system is described in the
following section. Computer vision methods are described in section 3. Object
identification, search and analysis techniques are described in section 4. The
NIST performance evaluation of the SUNAR system is in section 5. State of
the art is situated at the beginning of each section. The paper is concluded in
section 6.

2Sy emD gn

Although there are many multi-camera surveillance systems [10,7,12,13], we be-
lieve our approach outperforms the others, because those described in literature
were not evaluated successfully [10,12], while those in praxis make many simpli-
fying presumptions (e.g. trac monitoring). Moreover, in there is no need for
a central or primary module [7] or some special hardware such as camera sen-
sors [13]. Moreover, it is able to derive various useful information concerning the
entire area under surveillance.