A team of academic and industrial partners across Europe have demonstrated a prototype camera system dubbed Smart Eyes that analyses recorded data in real time and then zooms in on unusual scenes.
The system consists of a fixed surveillance camera that covers a certain area, and two ultra-active stereo cameras.
Marina Kolesnik, a scientist at the Fraunhofer Institute for Applied Information Technology, a SEARISE research partner in Sankt Augustin, said the system’s software mimics the processing in a human’s visual cortex. It begins by assessing the degree of movement for each pixel in the scene and picking out the most active areas.
Kolesnik said over a certain period of time it will learn typical motion patterns and store this as normal. So when something abnormal happens in a scene, such as a supporter jumping out of the stands and behind a fence, she said, it will identify it as a salient event. The active stereo cameras will then zoom in.
Kolesnik sees applications for this ranging from security cameras at football stadiums to city streets, underground metro systems and airports.
‘There is a huge amount of surveillance cameras around, taking in a huge amount of information and this information is impossible to view in its entirety,’ she said, estimating that up to 95 per cent of video recorded and stored is not significant or interesting.
‘This system is capable of selecting the most significant events in real time and then fixating those events and taking a higher-resolution record of only part of the scene,’ she added.
The SEARISE group is currently looking for commercial partners.
UK productivity hindered by digital skills deficit – report
This is a bit of a nebulous subject. There are several sub-disciplines of 'digital skills' which all need different approaches. ...