A centroid matching approach for real time foreground segmentation and occlusion handling

Demand and need in public place security has been raised in recent years. There is a wide scope of work to make a full automated real time video surveillance system to make a public place much safe and secure. The outcome of the CCTV and other digital surveillance can be enhanced by using real time foreground segmentation and occlusion handling technique. The outcome is used to analyse the suspicious activity and objects having important concern in public place security. The video frame foreground extraction and object prediction with occlusion is one of the major challenges in real time. Researchers are working on some novel foreground extraction techniques facing occlusion like major challenge in foreground feature extraction. In this paper a new parameter centroid matching has been proposed to extract foreground using k-mean clustering. It helps to track moving, occluded and unattended object in real time video frame.