Computer Vision and Pattern Recognition

Video analytics

There are now large networks of CCTV cameras collecting great amounts of image data, many of which possibily deploy Pan-Tilt-Zoom (PTZ) controllable cameras. A multi-camera and multi-sensor system has potential both for gaining improved imaging quality and for capturing more relevant details.

However, such systems a system can also cause overflow of information and confusion if data content is not properly analyzed, possibly in real-time. Video Analytics is the emerging technology where Computer Vision and Pattern Recognition techniques are used to filter, process and manage such huge amount pof data for security and intelligent monitoring, providing both low and high-level understanding of video streams.

PAVIS masters some of the cutting edge technologies in this area, which exploit recent findings in machine learning, statistics and differential geometry, and adapt them to challenging tasks, which comprise people detection, tracking and re-identification across non-overlapping cameras, activity recognition, crowd and group beahviour analysis, intention prediction. The robustness of our solutions is grounded on the many collaborations with industrial partners, which allow us to investigate what is really needed for the surveillance market.