Parallelization of Video Surveillance for GPGPU

This paper addresses the problem of Automated Video Surveillance (AVS), which involves automatically analyzing surveillance videos to detect suspicious or otherwise interesting activity. Interest in AVS is rapidly growing due to its wide range of applications, such as in homeland security, security for important buildings and shopping malls, traffic surveillance in cities and detection of military targets, etc . Video surveillance algorithms represent a class of problems that are both computational and data intensive. Obtaining the desired frame processing rate of 24-30 fps (frames per second) for such algorithms in real-time is one of the most important challenges faced by developers of video surveillance algorithms.In this paper, we implemented a video surveillance algorithm on a GPU and achieved the frame procesiing rate of 42.6 frames per sec which is higher than the required frame processing rate (> 25-30 frames per sec). We achieved an overall speedup of 12X on a high end commercial graphics processor (GTX 280). The GPU used is easily available as a graphics card and could be incorporated into most modern computers with small addition in the price.unpublishednot peer reviewe