Multi-view access monitoring and singularization in interlocks Luigi Di Stefano Federico Tombari Stefano Mattoccia Matteo Balasso Department of Electronics Computer Science and Systems (DEIS) Advanced Research Center on Electronic Systems (ARCES) University of Bologna {luigi.distefano, federico.tombari, stefano.mattoccia}@unibo.it matteo.balasso@studio.unibo.it http://vision.deis.unibo.it Paper ID 50 Abstract We present a method aimed at monitoring access to in- terlocks and secured entrance areas, which deploys two views in order to robustly perform intrusion detection and singularization. The main contributions are represented by an original approach to perform background subtrac- tion, which is particularly robust against sudden illumina- tion changes, shadows and photometric distortions, and by the use of a feature extraction and classification approach which allows to reliably determine an estimation of the number of people currently occupying the monitored area. Our system is designed to operate in very small interlocks and can work in a substantially unstructured environment. 1. Introduction Monitoring access to interlocks and secured entrance ar- eas, such as revolving doors, is a very important task in many security applications. The aim of the task is twofold: first of all, detecting the presence or absence of people in the monitored interlock, secondly allowing only one per- son at a time to be present in the interlock (singulariza- tion). Singularization is needed to avoid two (or more) peo- ple simultaneously crossing the gate (piggybacking) or an unauthorized person crossing the interlock other than the authorized one (tailgating). Most solutions deploy sensors such as weight controllers, light barriers and ultrasonic de- vices [9, 8] which are mounted inside the interlock walls and floor. However, these systems are generally expensive and not accurate enough. Furthermore, they are often un- practical for maintenance. Therefore, the use of simple and cost effective video- based approaches is gaining increasing attention in the se- curity industry. These approaches usually rely on back- ground subtraction to detect the presence and estimate the number of people in the interlock. However, it is difficult to perform reliable and accurate background subtraction in real environments, which are characterized by unstructured backgrounds (typically, the floor of the gate), sudden illumi- nation changes, presence of shadows. Commercial systems, such as Smart Airlock Control System (SMACS) by Fast- com Technology, deal with this issue by heavily structuring the working environment, i.e. by deploying markers on both the interlock floor (e.g. by means of a chessboard patterned carpet) and walls (e.g. by means of patterned stripes), and by requiring a specially aligned illumination to avoid shad- ows (i.e. by means of fluorescent tubes disposed in a rectan- gular arrangement). Another difficulty arises from the size of the interlock, that can often be very small. This con- strains the positioning of cameras and generally forces the monitoring system to work with very small field of views, which often do not allow the person to be seen entirely into one single view. The proposed approach exploits two views to accurately and robustly perform presence detection and singularization in small interlocks. The only modification to the environ- ment required by the system consists in installation of a tiny stripe of reflective material around the borders of the inter- lock floor. This is not needed for presence detection, but helps improving the feature extraction process required for singularization. Our approach is based on two main stages, which repre- sent also the two main contributions of this paper. On one IEEE Fifth International Conference on Advanced Video and Signal Based Surveillance 978-0-7695-3341-4/08 $25.00 © 2008 IEEE DOI 10.1109/AVSS.2008.22 140 IEEE Fifth International Conference on Advanced Video and Signal Based Surveillance 978-0-7695-3341-4/08 $25.00 © 2008 IEEE DOI 10.1109/AVSS.2008.22 143 IEEE Fifth International Conference on Advanced Video and Signal Based Surveillance 978-0-7695-3341-4/08 $25.00 © 2008 IEEE DOI 10.1109/AVSS.2008.22 143