A-LUX AI video analytics
WhatsApp
Workers in hard hats and hi-vis vests in a workshop with detection boxes
Workplace safety

PPE detection: hard hat and hi-vis vest monitoring by camera

The system watches workshop and yard cameras and notices anyone without a hard hat or a hi-vis vest. The responsible person gets a photo with place and time in Telegram while the violation is still happening.

Updated 17 November 2026 · A-LUX

How hard hat and vest detection works

Cameras stream video to a server in your server room. A neural network finds every person in the frame and checks head and torso: is there a hard hat, is there a hi-vis vest. If the violation lasts several seconds, the system takes a snapshot and sends it to Telegram.

Short events are ignored: taking the helmet off for a second to wipe your forehead is not an alert. Time thresholds and zones follow your safety rules.

Where to enable it

A zone polygon is drawn for each camera. People outside the polygon, in a locker room or corridor, are not checked.

What managers get

WhatHow it looks
AlertPhoto with a box, camera, area, time, type of violation
LogAll violations filtered by workshop, shift, date
ReportDaily and weekly summary in Telegram
AnalyticsWhere and when violations happen most often

Which cameras work

Any IP camera that streams over RTSP or ONVIF. Indoors 4 MP is enough when a person takes at least 1/10 of the frame height. In tall halls, from 12 metres, we mount cameras lower on columns or use narrower lenses. Existing cameras are tested on your recordings first.

Why not cloud or a boxed product

Video never leaves your site: recognition runs on a local server and only violation snapshots go out. We build the system ourselves and fine-tune the models on footage from your plant: your workwear, your helmets, your lighting.

PPE detection runs on the same cameras as face-based time tracking, idle time monitoring and crane danger zone.

Questions and answers

Does it detect a helmet worn over a hood or winter hat

Yes. The model is fine-tuned on footage from your site, including winter clothing. A hood without a helmet is flagged.

How many cameras are needed

Those that already cover dangerous areas. Usually 30-50% of plant cameras: workshops, warehouse, loading.

Can it check goggles, gloves, safety boots

Goggles and gloves are recognised when the camera is close. Boots are hard to see on wide shots, so we check them at the workshop entrance.

Is internet required

Recognition works offline. Internet is only needed to send snapshots to Telegram and for remote support.

We test it on footage from your cameras. For freeSend 2-4 hours of video from the gate or a workshop and we will show what the system catches on your own footage. Then a 30-day pilot.
Message us on WhatsAppAI video analytics

Other modules

AI video analyticsAI video analyticsFace recognition time tracking: timesheets without cardsTime and attendanceFace recognition access control with turnstilesAccess controlEmployee idle time and inactivity monitoringProduction analyticsCrane danger zone and man-down detectionIndustrial safetyIndustrial video analytics: pricing and stagesPricing and stages