Handling Occlusions in Monocular Surveillance Systems

Recent advances in computing machines and the availability of inexpensive vision sensors have paved the way for development of real-time imaging systems. Smart systems with a single fixed camera are often deployed for the task of outdoor surveillance. However, such systems are challenged by occlusions caused by interactions of foreground and background objects in the scene. In this paper, we propose an effective scheme for disambiguating such cases of occlusions and for the detection of entry and exit of objects using list based intelligent reasoning. The proposed methodology employs a statistical background model to identify foreground regions followed by smart tracking of the objects by using their color distribution and motion history. The present implementation runs at 7.5Hz while operating on color images of 320x240 resolution.

• Type: Conference
• Year: 2004
• Where: ICVGIP
• Authors: Prithwijit Guha, Nisarg Vyas, Amitabha Mukerjee, K. S. Venkatesh
• Keywords: ICVGIP 2004, ICVGIP 2008, Inexpensive Vision Sensors, Real-time Imaging Systems, Single Fixed Camera