Use case · Safety
Forklift–pedestrian detection
The highest-severity risk in most warehouses and plants, watched continuously by the cameras already on the ceiling.
The problem
Why it keeps happening.
Forklift strikes are among the most severe and most litigated industrial accidents, and near-misses happen daily without ever being recorded. Spotters, mirrors, and floor markings depend on humans being alert every minute of every shift, and safety teams only learn about close calls when someone reports one.
Most facilities already have camera coverage of exactly the intersections and aisles where conflicts happen. What is missing is something watching those feeds continuously and consistently.
The rule
Written the way you'd say it.
“Alert when a person comes within three meters of a moving vehicle in an active zone.”Compile it in the Rule Playground
How it runs
- 01
Connect the cameras you have
FloFrame connects to existing fixed cameras over ONVIF/RTSP: the aisle, dock, and intersection views you already own. No new cameras, no rewiring.
- 02
Define zones and the rule
Mark travel lanes, pedestrian walkways, and mixed zones on each camera view, then write the policy in plain English, distance thresholds, zones, and shift windows included.
- 03
Detect and alert in milliseconds
On-site GPU inference tracks people and vehicles simultaneously and fires the moment the rule is breached: to floor displays, radios, or the supervisor on shift.
- 04
Seal the evidence, trend the risk
Every event lands in FloFusion as a sealed, timestamped clip with an explanation. Near-miss trends by aisle, shift, and driver feed your safety reviews.
Measure
What's measured.
- Near misses per aisle, intersection and shift
- Closest approach between people and moving vehicles
- Time from alert to acknowledgement
What it’s worth
- Fewer struck-by incidents and the injuries, downtime, and claims that follow
- Near-miss data you have never had: intervene before the incident, not after
- Sealed evidence for insurance claims and disputes
- Traffic-flow insight that often pays for the deployment on its own
Estimate your ROI for your own operation, in about two minutes.
Comply
The evidence.
- A clip and frame of the event, sealed with the device’s own key the moment it happens
- The rule that fired, the zone and camera, and the time
- The vision-language model’s verdict on the alert, with its reason
- Who saw it, who acted, and when, in the audit log
Where it applies
Related industries.
Common questions
Does this work with cameras mounted high on the ceiling?
Yes. Detection works from the high, angled views typical of warehouse CCTV, and each view is checked against your own footage before its rule goes live.
Link to this questionCan alerts reach the forklift driver in the moment?
Yes. FloFrame can switch a floor beacon or siren on site through HTTP relays, Modbus TCP, ONVIF relays or MQTT, while FloFusion notifies supervisors and escalates if nobody acknowledges.
Link to this questionMore use cases
The use-case librarySafety
PPE detection
Helmets, vests, gloves, and eyewear verified continuously, not just when a supervisor happens to be looking.
Safety
Restricted zone monitoring
Machine cells, energized areas, and exclusion perimeters enforced automatically, every minute they matter.
Safety
Fall & man-down detection
When someone goes down, the response clock starts: detection in seconds, not whenever a colleague walks by.
See it on your own cameras.
A pilot on the cameras you already own, with success criteria agreed up front.