Single-Chip Design for Intelligent Surveillance System

Abstract

During recent years, a digital surveillance system has gained more popularity. Object recognition plays a critical role in an intelligent surveillance system. It mainly consists of several vision-based techniques such as segmentation and tracking. However, these techniques are still challenging. In this paper, an intelligent surveillance system has been proposed on the basis of single-chip design. An algorithm development and its hardware implementation are well addressed. Our algorithm accomplishes tracking of moving people through successive frames without using boundary box of object or color cues. Although there is occlusion due to gathering of people and other foreground objects, still the proposed algorithm can deal with such situations. The developed hardware architecture involves several tasks including foreground detection, sliced connected component labeling, object grouping, and object tracking. We use parallel processing on object labeling architecture and give the upper and lower image label values simultaneously. Object grouping and object tracking are also manipulated to improve the performance. The whole system is designed using single-chip solution. It is implemented by TSMC 90-nm library with 18.71-K logic gates, 92.288-kB on-chip memory, and 11.4037-mW low power consumption.