Intruder Recognition Security System Using an Improved Recurrent Motion Image Framework

Abstract

In this paper, we present an extension to the Recurrent Motion Image (RMI) motion-based object recognition framework for use in the development of an automated video surveillance Intruder Recognition Security (IRS) System. We extended the original object classes of RMI to include four-legged animal (such as dog and cat), and various enhancements are made to the object detection and classification algorithms for better object segmentation, error tolerance and recognition accuracy. Under the new framework, object blobs obtained from background subtraction of scenes are tracked using region correspondence. In turn, we calculate the RMI signatures based on the silhouettes of the object blobs for proper classification. The framework functions as the core of the IRS System to provide intruder recognition function and to reduce nuisance alarms since the system is capable of differentiating different category of objects in the surveillance area. A recognition rate of approximately 98% (40 out of 41 moving objects in the experiments were correctly classified) was achieve in our tests based on several real world 320 ×240 resolution color image sequences captured with a low-end digital camera, and also on the PETS 2001 dataset. Thus, indicating the applicability of the new RMI framework to minimize nuisance alarms in an IRS System.