Video Surveillance and Biometric Technology Applications

Summary form only given. The paper refers some recent experiences on Video Surveillance applications in public spaces, with particular attention to the airport scenario. After a brief discussion on the main open problems and the level of performance that is currently achieved by vision technology, the issue of integration with biometric technology is addressed. Some preliminary results of an Italian national research project are also referred, with an experimental demonstration to simulate the typical problems of security access control in an airport scenario. The MULTI-TRUST project was intended to study the main market trends and innovation technology in the airport field, to study the potential of 3D face recognition technology, and solutions of Intelligent Video Analysis for people tracking. The processing scheme is a layered architecture with different modules at the level of info-processing, entity fusion and high-level event detection and classification. A significant contribution came also from the development of an interactive station for people enrollment, check-in booth functions and passenger identity verification. Multi-Biometric combination has proved to be quite a promising approach to achieve a reliable person identification, including different biometric functions like fingerprint with face features (possibly 3D) and other personal identification data. Moreover Video Analysis is a key feature of new advanced Video Surveillance Systems to provide an intelligent support to security operators to highlight the critical situations while reducing the number of false alarm events. Further technological improvements are required in this field, since the current level of performance is not always appropriate for the severe requirements of the users. In such a context 3D modeling represent the most effective way for people/object detection and classification. Also the experience of MULTI-TRUST has demonstrated that Computer Vision technology is not yet reliab- le enough to manage very crowded scenes and multiple occlusions in variable light conditions. Robustness of processing remains a challenge. Finally, Behavior modeling and classification represents the new frontier of research with an increasing potential of security applications.