The Ratio of False to Valid Alerts is Key

Published Oct 21, 2009 00:00 AM
PUBLIC - This article does not require an IPVM subscription. Feel free to share.

While false alerts rightfully receives the majority of attention in video analytic performance critiques, these systems must also pass another test - that of balancing the number of false and valid alerts.

VideoIQ provides an example of this in their analysis of baggage left behind problems. Baggage left behind suffers from inevitable false alerts as normal people in public areas temporarily set down their baggage. Additionally, and equally importantly, the likelihood that a bag is really a bomb is extremely low, even in a high risk areas.

The combination of high false alerts (numerous per day) and low actual threats (once per decade or longer?) makes the probability that any actual alarm being a valid one to be worse than 1 in 1,000 (easily).

Academic research [link no longer available] (as well as common sense) shows that such circumstances significantly reduces the effectiveness of such systems.

This ratio can equally be applied to other types of video analytics where the event has a very low probability of occurring.

Minimally, dedicated monitoring services are key to handling this as multi-tasking operators would likely abandon such systems as distracting and unproductive.