Configuring Camera-Based Smoking Detection at a Workplace
Video-based smoking detection is possible where cameras distinguish relevant objects and actions. Configure it for specific zones and send detected episodes for human review. At Busted, we start with source-image assessment and an agreed event definition.
Why a hand movement is not enough
People raise hands to their faces when using phones, drinking, adjusting glasses or smoking. Responding only to a gesture may create too many ambiguous signals. We assess combinations of features and their sequence over time.
Distance, object size, lighting and occlusion matter. A wide factory view may show presence well but lack detail for this action. Assess these limitations using examples.
Defining the monitoring zone
The business specifies where and when monitoring is needed. Agree boundaries, operating hours and event wording. Different areas may need different rules, so transferring a configuration to another camera requires validation.
Discuss whether to count confirmed episodes only or also measure duration. Duration needs stable start and end definitions and separate accuracy assessment.
What the result should provide
A useful event card includes camera, location, time and a short frame sequence. Reviewers can see the context and record an outcome. Reports separate confirmed and rejected cases so managers use verified data.
The pilot needs smoking examples and similar everyday actions. Comparing them establishes suitability for the site. Then assess misses, unnecessary alerts and review time.
What to send for assessment
Provide monitored-zone descriptions, original clips and examples of situations worth flagging. Busted helps define observable features and validate the scenario before broader rollout.