Person re-identification (ReID): following one journey across every camera
What ReID is, why it matters more than single-camera tracking, why it is hard in real sites, and how Eagle Vision’s approach links one person across cameras without needing a face.
One person, many cameras
Most video analytics see the world one camera at a time. A person walks out of camera 3, and when they appear on camera 11 they become someone new. For many questions that matter to security and operations teams, this is exactly where the answer gets lost.
Person re-identification, or ReID, solves this. It recognises that the person on camera 11 is the same person who left camera 3 a minute ago, and joins the two sightings into one journey. It does this from overall appearance, such as clothing, shape and carried items, so it works from behind, from a distance and when the face is not visible.
| Single-camera tracking | Face recognition | ReID | |
|---|---|---|---|
| Answers | Where is this person in this view? | Who is this person? | Where did this person go next? |
| Needs a face | No | Yes, clear and frontal | No |
| Works across cameras | No | Yes, if the face is seen | Yes |
| Identity | Anonymous | Named | Anonymous by default |
Why ReID matters
- Investigations in minutes: click a person once and see every camera they passed, in order, with times.
- Real journeys in malls and stores: entrance to zones to cashier to exit, without counting the same visitor twice.
- Unique visitor counts across a site with several entrances and floors.
- Loss prevention: follow an item-carrying person from the stockroom to the exit.
- Safety: confirm a lone worker reached the muster point, or find a lost child from one photo, with the right consent and process.
- Time between zones: how long it really takes from the gate to the loading dock, or from the queue to the counter.
Why ReID is hard in real sites
Research benchmarks look clean. Real sites are not. The same person looks different from a ceiling camera and from a corridor camera, under warm indoor light and harsh sun, half hidden behind a shelf or a crowd.
- Viewpoint and pose change from camera to camera.
- Lighting and color shift between indoor, outdoor and night footage.
- Uniforms: in factories, restaurants and sites, dozens of people wear exactly the same clothes.
- Occlusion and crowds hide parts of the body.
- Low resolution on wide, older or distant cameras.
How Eagle Vision solves it
Our ReID was built for the sites our clients actually run, not for benchmark videos. Instead of relying on appearance alone, it combines several signals and only joins two sightings when they agree.
- 01Appearance signature
Each person gets a compact appearance signature built from the whole body and what they carry, robust to pose and lighting changes.
- 02Site map and timing
The system learns which cameras connect to which, and how long the walk between them takes, so impossible matches are ruled out.
- 03Movement and context
Direction of travel, speed and the zone a person left from are combined with appearance, which keeps uniformed staff apart.
- 04One journey
Confirmed sightings are joined into a single timeline you can search, replay and export as evidence.
ReID across industries
| Industry | What ReID adds |
|---|---|
| Retail and malls | Full shopper paths, cross-zone dwell, unique visitors and true conversion. |
| Security | Search by example across all cameras and build an incident timeline in minutes. |
| Factories and warehouses | Who entered a restricted zone and where they went next; time from gate to dock. |
| Construction | Headcount per zone and confirmation that everyone left a danger area. |
| Events and public spaces | Locate a lost child or a person of concern from a single photo, under an agreed procedure. |
Privacy and governance
- Appearance signatures are numbers used for matching, not photos, and are kept only as long as you decide.
- Journeys are short-term by design; matching across days needs face recognition, which is a separate decision.
- Role-based access and an audit log record who searched for whom, and why.
- Processing can run fully on-premise so video never leaves your network.
Questions we often hear
Is ReID the same as face recognition?
No. ReID links sightings of the same person across cameras from overall appearance and gives them an anonymous ID. Face recognition identifies who a person is and needs a clear view of the face.
Does ReID work if people wear the same uniform?
This is the hardest case. Eagle Vision combines appearance with the site camera map, timing and movement, which is what keeps people in identical uniforms apart.
Can ReID follow a person across different days?
ReID is designed for journeys within a visit, because clothing changes between days. Recognising the same person on another day requires face recognition.