Monday, August 4, 2008

Video Recognition System

There is an interesting article on that describes a technology that might solve one of the major problems associated with observing the outputs of multiple video cameras in complex video surveillance system, operator attention fatigue. A human operator can only watch video monitors for a relatively short period of time before his attention begins to wander. The larger the number of cameras involved, the shorter the effective attention span.


Computer Vision


The computer technology developed by BRS Labs is supposed to allow for a conversion of a video signal to machine language. This allows the software to check for changes in the field of view and learn what changes are ‘normal’ for a particular location. When something abnormal occurs in the field of view of the system, an alarm is sounded, requesting human intervention.


This should allow a human operator to more effectively monitor a large number of cameras. Rather than try to continuously scan a number of screens, the operator would be able to concentrate on those where the computer reports unusual changes in the visual environment. This should also allow for a reduction in the number of video monitors required at the security monitoring stations.


System Effectiveness


According to the article BRS Labs expects a 3 to 1 false positive rate. No information is provided on the false negative rate, to my mind a much more serious measurement of system effectiveness.


The system won’t be widely available for a few months. This means that there is not a large scale reliability history on the equipment or software system. A facility with a large number of cameras might still find this a worth while system to try during the early stages of its deployment. The problems of video monitoring are so extensive that even a relatively small incremental improvement might be very worthwhile.

No comments:

/* Use this with templates/template-twocol.html */