Dual Background Segmentation Based Abandoned Object Detection in Video Surveillance Systems

Abstract: Detection of unattended object is one of the important tasks in video surveillance system. This paper describes the unattended object detection based on dual background segmentation approach. The proposed system consists of preprocessing model which includes background segmentation, shadow removal and brightness balancing techniques. The system uses two sets of background i.e current background and buffered background for the detection of foreground blob. The framework of the system is based on approximate median model. The stationary object is detected and based on the threshold value the alarm is raised after detecting the unattended object. The system is tested on a video of resolution 320�240 pixels containing all types of objects. i.e. moving objects, static objects and abandoned/unattended objects.

Keywords: Video Surveillance, Image Pre-processing, Dual Background Subtraction, Tracking, Abandoned Object.