Definition
Video analytics is the automatic processing of camera images to detect events and collect numerical data about what happens in the observed area. The output is not footage but a record of events and indicators that can be compared over time and between areas of a plant.
What is actually measured
The most common indicators are the number of events of a given type, dwell time in an area, traffic intensity by time of day, counts of people and vehicles, and the distribution of events across a floor map.
Value comes from comparison over time. A single alert says little; the information that one area produces five times more events than the others points to a place where work organisation should change.
Analytics versus detection
Detection answers the question of whether something is happening now and triggers a response. Analytics answers how often and where, and serves decisions rather than alarms.
The distinction has technical consequences: detection must run in real time and locally, while analytics can compute more slowly and on data stripped of imagery.
What analytics will not do
It will not explain the cause. It will show that a given walkway sees frequent pedestrian-forklift encounters, but a person has to establish whether the problem lies in signage, delivery scheduling or rack layout.
It also does not replace talking to the crew. Data shows the effect; the cause is usually known to the people working there and asking them is enough.
Frequently asked questions
Does video analytics require recording?
Not always. Indicators can be collected without storing footage, keeping recordings only for events that need verification.
How long must data be collected before it means anything?
Usually a few weeks. A shorter period shows random fluctuation rather than repeatable patterns.
Can analytics data be exported?
Yes, through an API or reports. This matters because conclusions are usually cross-checked against data from other plant systems.