Object Tracking over Multiple Uncalibrated Cameras Using Visual, Spatial and Temporal Similarities - Advanced Concepts for Intelligent Vision Systems - page 166

Object Tracking over Multiple Uncalibrated
Cameras Using Visual, Spatial and Temporal
Similarities

Daniel Wedge , Adele F. Scott, Zhonghua Ma, and Jeroen Vendrig

Canon Information Systems Research Australia
[e-mail]
http://www.cisra.com.au

Abstract. Developing a practical multi-camera tracking solution for au-
tonomous camera networks is a very challenging task, due to numerous
constraints such as limited memory and processing power, heterogeneous
visual characteristics of objects between camera views, and limited setup
time and installation knowledge for camera calibration. In this paper, we
propose a unified multi-camera tracking framework, which can run online
in real-time and can handle both independent field of view and common
field of view cases. No camera calibration, knowledge of the relative po-
sitions of cameras, or entry and exit locations of objects is required. The
memory footprint of the framework is minimised by the introduction
of reusing kernels. The heterogeneous visual characteristics of objects
are addressed by a novel location-based kernel matching method. The
proposed framework has been evaluated using real videos captured in
multiple indoor settings. The framework achieves ecient memory usage
without compromising tracking accuracy.

Keywords: distributed tracking, surveillance, real-time systems.

1 Introduction and Related Work

The large sizes of modern surveillance camera networks mean that it may not
always be possible for a human being to monitor every video stream in real-time.
This presents the need for autonomous camera networks that can extract video
content for either human users or higher-level autonomous processing. It is de-
sirable in industry applications for these networks to operate in a decentralised
manner in order to minimise setup time, for improved robustness and scalability,
and so that in the case of active camera networks each camera may act as an
autonomous agent. The objective of multiple camera tracking, or, multi-camera
tracking, is to determine correspondences between observations of real-world ob-
jects seen by multiple cameras after object detection and single-camera tracking
have been performed. We consider the case where cameras are uncalibrated and
there is no knowledge of the network topology. Multi-camera tracking methods

The author is currently with the University of Western Australia.