Multi-object tracking
Multi-object tracking is the computer vision process that detects every person, vehicle or object in a camera view and follows each one from frame to frame as the same subject. It is what lets a system say that a person crossed a line, stayed in a zone or passed through a door, rather than that pixels changed.
A tracker keeps an identity for each subject through brief occlusions and re-entries, so a person who walks behind a pillar comes out as the same track. Every rule on the camera, line, zone, dwell time or count, is evaluated against those tracks.
The identity is a track number, not a person's identity. Tracking does not recognize faces; it follows shapes and motion, which is why detections built on it can run without identifying anyone.
On Camzify multi-object tracking underlies every detection and the patrol round, and pairs with AI attribute extraction to attach a plain description such as clothing color to a track when that feature is enabled.