Human classification
Detects people, not motion; shadows, leaves and headlight reflections do not trigger alarms.
ProVife Human classification kicks in when Person is selected as the detection object in zone and line scenarios. The AI recognizes people in the image, assigns each a tracking ID and responds only to these targets. Trees swaying in the wind, rain, insects, cloud shadows or the headlights of a passing vehicle do not produce events. This eliminates the unnecessary notifications that pile up overnight in perimeter security, so the monitoring center deals only with real human movement. When alarm mode is armed, the same event can also directly start smart building scenarios such as sirens, lighting and speaker announcements.
How it works
Grab the camera’s current image with Capture frame and draw the area to monitor by dragging on the image.
Select Person from the detection object list; in this scenario ProVife treats only people as targets.
Use the confidence threshold to set how certain the AI must be about a person.
Each detected person gets a tracking ID and is tracked with the same ID as long as they stay in the area.
If the dwell time is zero, an event is created the moment the person enters the area; if a time is set, when it elapses.
Test the settings on live video with the Try button, then save the scenario to the server.
Settings
| Detection object | With Person selected, the scenario responds only to people; the same list also includes vehicles, animals and specific objects. |
|---|---|
| Area | The inspection rectangle drawn on the image; motion outside it is not evaluated. |
| Confidence threshold | A lower value catches more people; a higher value accepts only clearly visible people. |
| Dwell | How long the person must stay in the area before an event is created; filters out brief passes. |
| Repeat interval | The minimum time that must pass between two notifications for the same person. |
| Analytics hours | The schedule that sets which days and hours of the week the scenario runs. |
What it produces
- Zone intrusion event; the record includes person as the object class and the person’s tracking ID.
- Event image with the person marked by a red box and the scenario area by a yellow line.
- Live number of people in the area on the smart panel.
- Time the person spent in the area, in seconds.
What you can do with the event
- A ready-made scenario that switches alarm mode to triggered when a person is seen while Away mode is armed.
- Turning on yard or parking lot lighting for a set time when a person is detected at night.
- Sending the event image with your own description to a WhatsApp group.
- Showing the relevant camera in a pop-up window on the operator screen and on the floor plan.
- Publishing the event via MQTT to the building management system.
Where it is used
Industrial facilities
After hours, only human movement along the chain-link fence triggers alarms, so windy nights bring no flood of notifications.
Residential complexes
A person seen in common areas at night reaches the security booth as a notification with an image.
Construction sites
A person entering the material yard outside working hours is met with a siren and an announcement.
Warehousing
In an indoor warehouse, every person seen after the shutters close is logged.
Camera placement
- Look from about 10–16 ft high, at a distance where the person takes up at least a few tenths of the image height.
- For night use, light the area sufficiently with infrared or white light.
- Avoid angles facing directly into the sun or headlights.
- Draw the area only over the zone where people should not be; leave the road and sidewalk out.