Core Definition
MTMC (Multi-Target Multi-Camera Tracking) is a computer-vision technique that joins movement captured across multiple cameras, tracking many objects continuously across the whole store. One camera's blind spot is covered by another, producing unbroken pathways from entry to exit.
Why one camera is not enough
No single camera captures an entire store. Behind display shelves, around corners, at the end of an aisle, every camera has a blind spot, and a customer who disappears from one camera reappears in another. Viewed camera by camera, paths remain fragmented.
MTMC stitches these fragments together. It projects each camera's view onto a single coordinate system, the store floor plan, and uses anonymized features to connect the same object across camera boundaries. The result is a path that runs through the whole store, such as "entry → display A → display C → checkout."
Use in store analysis
An unbroken path is the foundation for zone-level analysis. Questions like how much movement happens between two zones, which route counts as the main path, and where foot traffic drops off are only accurate once MTMC is in place.
The effect grows with the size of the space. In places that need dozens of cameras, such as exhibition spaces, hypermarkets, and logistics facilities, MTMC summarizes the flow of the entire space into a single map.
How SAAI Uses It
DEEPINGSOURCE's Spatial AI uses MTMC to integrate multiple camera views onto a single floor-plan coordinate system. Because it reads paths, dwell, and movement between zones against the whole store rather than individual cameras, it enables analyses such as saai insight's main paths and traffic between zones. All tracking runs on data anonymized at the point of input.
See it in action