Construct-AI-primary PPE Detection
Overview
Computer vision-based detection and classification of personal protective equipment compliance on construction sites. Analyses video feeds or photographs to identify persons and check PPE against organisational requirements. Primary agent: CV/IoT Integration Specialist.
Steps
- Image Preprocessing → Prepare camera frame or photo: resize, normalise, enhance
- Person Detection → Identify all persons in the image using object detection model
- PPE Analysis → For each detected person, analyse: hard hat (on/off), high-vis vest (on/off), safety boots (on/off), eye protection, hearing protection, gloves, fall protection
- Compliance Assessment → Compare detected PPE against required PPE for the zone or activity
- Violation Documentation → Document any violations with person location, timestamp, missing PPE
- Alert Generation → Generate real-time alert for serious or immediate danger violations
- Data Logging → Log all PPE detection results for trend analysis
Success Criteria
- Detection accuracy > 90% for each PPE type in good conditions
- False positive rate < 5%
- Violations documented with evidence
- Real-time alerts generated for serious violations
Common Pitfalls
- Poor lighting conditions → Affects detection accuracy
- Occlusion → PPE partially obscured or person blocked
- Camera angle → PPE not visible from certain angles
- Similar objects → Non-PPE objects confused with PPE
Cross-References
safety-hazard-detection/SKILL.md— Hazard detection companionsafety-alert-system/SKILL.md— Alert generation