EURASIP Journal on Embedded Systems 
Volume 2006 (2006), Article ID 45758, 12 pages
doi:10.1155/ES/2006/45758

Customizing Multiprocessor Implementation of an Automated Video Surveillance System

Gary Wang, Zoran Salcic, and Morteza Biglari-Abhari

 Department of Electrical and Computer Engineering, University of Auckland, Private Bag 92019, Auckland 1142, New Zealand
Received 11 December 2005; Revised 4 July 2006; Accepted 12 July 2006

Recommended by Leonel Sousa

Abstract

This paper reports on the development of an automated embedded video surveillance system using two customized embedded RISC processors. The application is partitioned into object tracking and video stream encoding subsystems. The real-time object tracker is able to detect and track moving objects by video images of scenes taken by stationary cameras. It is based on the block-matching algorithm. The video stream encoding involves the optimization of an international telecommunications union (ITU)-T H.263 baseline video encoder for quarter common intermediate format (QCIF) and common intermediate format (CIF) resolution images. The two subsystems running on two processor cores were integrated and a simple protocol was added to realize the automated video surveillance system. The experimental results show that the system is capable of detecting, tracking, and encoding QCIF and CIF resolution images with object movements in them in real-time. With low cycle-count, low-transistor count, and low-power consumption requirements, the system is ideal for deployment in remote locations.