• Skills:

    • C++
    • Computer Vision

In the last years video-surveillance evolved from the massive use of analogical cameras to cameras that produce digital outputs, sometimes in high definition, and stream them over IP to allow remote surveillance. Internet brought brand new approaches to the video-surveillance world even if it has some limitations.

On the other side, the computer vision has grown a lot over the last years allowing to bring video content analysis (VCA) from the researcher's desk to the final user. VCA allows to automatically analyze videos and images to extract relevant information for the user: e.g. if a person is crossing a virtual or real border, alerting if a certain area is too crowd, the use of forbidden areas, and etc.

Actually, the vast majority of the VCA tasks are run on servers due computational load (e.g. searching for people identities) and only simple tasks (e.g. border crossing) can be demanded to remote device's processor .

SIBIRI wants to overcame this limit by focusing on the analytics on the edge functionality, and generally targeting the smart functionalities by developing devices with higher computational capabilities with respect those available now in the market and the capabilities of running VCA tasks on the device.

In this project we are teaming up with Infomob, who is the project's leader, and our role is that one to exploit our knowledge and competence in computer vision and machine learning to bring VCA on the SIBIRI's device.

This project is cofinanced by POR FESR 2007-2013 – Incentivo Ricerca Polaris BigData Image