Development of Low-Power Intelligent Video Surveillance System Using Embedded Computer

There has been a surge in criminal activity due to rapid urbanization, which poses a threat to community safety. Centralized video surveillance systems using computer vision have been introduced to monitor and respond to such activities. However, their implementation is limited by the substantial costs of both initial installation and maintenance. To tackle this issue, low-cost distributed embedded video surveillance systems have been proposed, utilizing low-power processors to minimize costs but with limited computational capabilities. To overcome this limitation, this study introduces an intelligent video surveillance algorithm that uses effective and simple algorithms, such as background subtraction and Frame-Skipping, to detect loitering, intrusion, and arson in embedded computing environments. As a result, a low-power distributed video surveillance system that can sustain 29 frames per second within the embedded system environment has been developed and experimentally verified.