Synthetic and privacy-preserving visualization of video sensor networks output

Claudia ARAIMO - intership student
Multimedia Communications

Date: -
Location: Eurecom

In the last years the spread of video cameras in surveillance systems has led to a video information overload. As a result human monitoring activity became more and more expensive or ineffective. Furthermore, such a large amount of data generates important concerns about privacy of recorded people. Here I propose a framework for multicamera videosurveillance system that is able to track people moving in a large site, to extract their trajectories and to project them in real-time on the map of the surveilled areas. Such a system allows a single human operator to effectively monitor human activities in a indoor environment by looking at a single monitor. Indeed my approach consists in synthesizing the output data from a video camera network into the building map of the surveilled areas. Having all the information condensed into such a unique picture allows the summarization of human activities without loss of useful information. Furthermore the privacy of recorded people is preserved since none of sensitive data is showed.