AI algorithm

Smoking and Phone Use Detection

AI detection for smoking and phone-use behavior in gas stations, factories, warehouses, hazardous-operation areas, and restricted work zones.

Request a Demo View Products

Smoking and phone-use analysis is used where a site has a specific behavior rule, such as a fueling area, unloading zone, workshop, warehouse aisle, or hazardous-operation area. It should be deployed as a reviewable early-warning tool, with the camera position and the event rule agreed before acceptance.

Practical Limitations

If the person is too far away, facing away from the camera, heavily occluded, or the hand is not visible, recognition stability decreases. The deployment should combine camera review, ROI rules, threshold tuning, and event review.

Algorithm deployment notes

Smoking and phone-use checks need close, reviewable views

Smoking and phone-use events involve small objects and short actions. Their practical accuracy depends heavily on image detail, head-and-hand visibility, lighting, and the duration of the observed action. These models should be deployed where a site can obtain a close, stable view of the relevant activity.

Prioritise a clear upper-body image

Entry points, waiting areas, workstations, dispensing islands, duty rooms, and controlled access points can provide better views than a distant overview camera. Check whether hands, face, head, and the likely object position remain visible when people turn, bend, carry tools, or pass one another.

Use persistence and review

Brief hand movements, pens, radios, food, protective equipment, or reflective objects may look similar to the intended action in some views. Configure a suitable observation interval, repeat rule, and review process instead of relying on a single low-detail frame for an automatic operational decision.

Capture enough surrounding context

An event should retain the camera, location, time, zone, detection type, snapshot, and if available a video reference. The surrounding scene matters because it shows whether the activity occurred inside a restricted or high-risk area.

Validate at the actual working distance

Test with representative staff positions, uniforms, tools, lighting, weather, and camera angles. Record the distance and conditions where the site considers the image sufficient for a reviewable event, and direct uncertain cases to human confirmation.

Related planning: Gas station video analytics overview. The alert procedure should be agreed with the site operations and safety team before use.

Project inquiry

Tell us your camera environment and required AI analytics.

For the first version this site uses a lightweight email-based inquiry flow. It avoids a database and keeps the English site simple until inquiry volume justifies a dedicated CRM integration.