الفهرس 
Techniques for Video Background subtraction
Future WorkOnly 14 pages are availabe for public view
 |  | from | 146 |  |  |
| | | from | 146 | | |
Abstract
One of the most important steps in any smart video surveillance systems is background removal. The background removal process facilitates identifying objects of interest mainly human beings. This important step in video processing is a key step in any smart video processing system and especially surveillance systems in crowded places like airports, train stations, shopping malls and residential areas where the need of security is highly recommended and where terrorist attacks is highly possible. This main step opens the door for many other video processing steps like objects classification, tracking, activity understanding and finally machine interaction. However, researchers are interested in solving this problem using different techniques and different platforms many years ago; they didn’t discuss how to remove the background from multiple videos in a concurrent way in real time. Many researchers handle one video background removal problem which is very costly in terms of computation and memory consumption. The need of a high processing power and a big memory is a must if we deal with multiple videos background removal that exists in smart video surveillance systems. The number of videos processed could be a real challenge and sometimes a system limitation if we use an ordinary CPU platform system; Instead we suggests in our research a Graphics Processing Unit (GPU) to exploit its massively parallel nature and their high concurrent processing power to accelerate the background removal process of multiple videos achieving real time display for all of them. One of the bottlenecks of using a GPU is the memory latency in transferring data to and from the GPU. In our research we introduce some optimizations principles to overcome that problem and due to these optimizations we succeeded to reduce the effect of memory latency to some extent and display many videos concurrently with background removed. The algorithm used in the background removal is a statistical approach model named Single Gaussian algorithm for background subtraction.