VICATS is a bi-national project led by Serbia (33.33%) associated with Belgian partners (66.67%). The VICATS partners are the Novid Sad Faculty of Technical Sciences, Traficon, the Image Processing and Interpretation (IPI) research group of the Ghent University, the University College of Ghent (Hogent) and Fitis-Ju. Visit www.ursusgroup.com, www.traficon.com, http://telin.ugent.be/ipi/drupal, www.hogent.be and www.fitis-ju.co.yu
Fitis-Ju is a major Serbian supplier of video surveillance systems (VSS), with a well-developed sales
network. Apart from national references in Serbia, Fitis-Ju has provided security equipment for many objects in Montenegro, Russia and Ukraine. So far, Fitis-Ju has provided complete VSS solutions for numerous large objects.
Project description
Contemporary video surveillance systems rely heavily on intensive human effort for detecting and assessing threats. Nowadays, the technology has reached a stage where mounting cameras to capture video imagery is cheap, but finding available human resources to watch the footage full-time is expensive and unreliable.
Besides being unable to maintain constant attention for long periods of time, humans cannot process a load of information contained in complicated scenes. In modern surveillance systems, automated solutions are sought instead of humans for continuously analysing the imagery. Intensive research in the fields of computer and video technology in the last decade has resulted in the development of clever algorithms that can automatically analyse video data and make decisions about the content. However, the existing algorithms can usually operate only in controlled, laboratory conditions due to the lack of robustness, or they are highly complex and require excessive computer power and processing time.
This project will focus on the development of robust and high performance computer vision algorithms for real-life automated video surveillance and traffic monitoring. The developed algorithms will be applied in a software module for video content analysis and integrated into operational video surveillance systems, as well as in newly installed systems. The target application is surveillance of critical traffic infrastructures, while the main focus will be on the tunnels. However, the results obtained can be applied to other critical parts of the transportation system.
One of the most important applications of video surveillance systems is traffic surveillance. The goal of
automated traffic monitoring is to create a completely autonomous system that gathers detailed information on traffic conditions 24-hours a day and deduces higher-level knowledge such as pedestrian detection, queue detection, vehicle classification, vehicle counting, average speed calculation, detection of violations of speed limits or other traffic regulations, accident detection etc. Such systems can be successfully applied to open roads, as well as tunnels and intersections.
The software for automated monitoring to be developed within this project will be based on the advanced computer vision algorithms that have emerged recently in the research community. The existing algorithms perform well only in laboratory conditions and cannot be directly applied to real-world situations. In order to be integrated into existing video surveillance systems, the most promising algorithms need to be identified and evaluated, and then modified and extended in order to satisfy the requirements of the application in question.
The existing video surveillance systems use predefined off-the-shelf hardware (cameras, sensors, etc.) and commercially available software packages for system integration. These software packages are responsible for bringing signals from all cameras and sensors in the system to one concentration point (the surveillance centre), where these signals can be separately viewed, recorded, etc. However, they offer no possibilities for any kind of intelligent signal analysis. That is why such software packages are designed as open-architecture systems, upgradeable with various specialised software modules. Separate modules need to be designed and devoted to specific applications such as traffic monitoring, airport security, etc., that can be integrated with open-architecture systems.
The developed video surveillance systems will reliably track moving objects such as pedestrians and
vehicles, and report their annotated trajectories to the threat assessment module for evaluation, thus making a surveillance system capable of performing tasks that were not possible with ordinary CCTV (closed circuit television) systems. The system is meant to be fully automated, without any need for a human operator to initialise it, and should work all the time and in all weather conditions.
Partners in this project form a strong consortium that can foster integration of existing theoretical knowledge and modelling and simulation abilities with deployment, testing and long-standing experience in the field of video surveillance systems. The combined expertise within the consortium ensures that the goals of the project will be reached and that new intelligent surveillance algorithms will de developed according to the needs of participants from the industry. 26/08.
