Multi-camera human behavior monitoring and unusual event detection.

Camera networks are already in wide use for surveillance purposes and traffic monitoring, and will be used for a growing number of emerging applications, such as elderly care, domotics, and security and safety. Current technological developments will result in large camera networks that record an enormous amount of video. The grand challenge is to come up with robust algorithms that automatically extract only the relevant information from the video streams. Current state-of-the-art computer vision algorithms, e.g., algorithms for fall detection of elderly persons in a camera based home monitoring system, perform poorly in real-life situation, especially under changing conditions (lighting, environment, ). Furthermore, processing all video on a central server is no longer desirable and needs to be re-placed by smart cameras that process the video themselves. The aim of this project is to develop novel techniques for long term activity monitoring and behavior analysis of people indoors using distributed smart camera systems. In this project we will explore and integrate two complementary ap-proaches: on the one hand multi-camera collaborative algorithms for occupancy mapping and tracking using many low-resolution cameras; on the other hand single-camera algorithms for detailed behavior analysis and action recognition using a few high-quality cameras.

Keywords:Human behavior monitoring, Unusual event detection, Surveillance, Video/image interpretation, Action recognition, Video monitoring

Disciplines:Applied mathematics in specific fields, Multimedia processing, Biological system engineering, Signal processing, Computer architecture and networks, Distributed computing, Information sciences, Information systems, Programming languages, Scientific computing, Theoretical computer science, Visual computing, Other information and computing sciences