AI Helmet, Vest & PPE Detection for Abu Dhabi Construction Sites

The mechanism is a multi-class object detector applied to each person bounding box inside a defined zone polygon. The detector picks the person; a secondary classifier or a multi-head detector then evaluates whether a helmet sits in the head region, whether a hi-vis vest covers the torso region, and whether other PPE classes are present. Confidence scores are produced per item, and a violation event is raised only when the per-item confidence is above a tuned threshold and the violation persists across a short temporal window — typically a few seconds — so a worker glancing down to adjust their helmet does not trigger a false alert. The realities that shape an Abu Dhabi deployment are these. First, summer heat means workers genuinely remove helmets in shaded rest zones; the zone polygon must distinguish the rest zone from the active-work zone, or the system will flood the report with false-positive alerts that the safety officer learns to ignore. Second, hi-vis vests come in different shades — yellow, orange, lime — and dust accumulation across a long Mussafah summer day shifts the colour signature; the classifier must be trained or fine-tuned against the colours actually worn on site. Third, low-mounted cameras at gate height create steep upward angles that defeat helmet-on-head detection; the camera must sit slightly above head height with a clear top-of-head view. Emirtech walks the site, sets the zones, and tunes thresholds against your own footage rather than relying on a vendor promise.