Computer vision identifies serious-injury risk by detecting, in camera video, the conditions that come before serious incidents rather than the incidents themselves. Typical detections include people walking in forklift lanes, close passes between trucks and pedestrians, missing required PPE, loads carried too high, blocked exits and people under suspended loads. Each event is logged with location, time and a severity category, so EHS teams can see where high-energy exposure repeats. Some systems also act in real time by switching warning lights or alerting operators. The output is exposure data, and people still need to review it and change controls.
The full guide is AI forklift safety systems. Accuracy depends on camera placement, lighting and how well the model was trained on scenes like yours, so test it on your own footage. Ask how severity is assigned, which event types are counted and how false detections are filtered before they reach a dashboard. Video of identifiable workers raises privacy duties; check retention, blurring and employee notice, covered in AI camera privacy. Pair the data with a serious-injury classification method such as the one in SIF exposure.