Safety Hazard Detection

Safety Hazard Detection

Overview

Computer vision-based detection of physical safety hazards on construction sites including open edges, unguarded openings, unshored excavations, blocked access routes, unsafe scaffolding, fire hazards, and hazardous material exposures. Primary agent: CV/IoT Integration Specialist.

Steps

1. Image Preprocessing → Prepare camera frame or photo for analysis
2. Scene Understanding → Identify the environment type and current work activities
3. Hazard Detection → Analyse for hazards: open edges without protection, unguarded openings, unshored excavations, blocked access or egress routes, unsafe scaffolding, fire hazards, hazardous material exposures, falling object risks
4. Severity Assessment → Classify hazard severity: Low, Moderate, High, Critical
5. Location Tagging → Record hazard location relative to camera and coordinates if known
6. Alert Generation → Generate immediate alert for high or critical hazards
7. Documentation → Log hazard with evidence photograph for follow-up

Success Criteria

• Hazard detection accuracy > 90% for defined hazard types
• Classification matches actual hazard severity
• Alerts generated for serious hazards without delay
• All detections documented with evidence

Common Pitfalls

1. Similar objects → Temporary materials confused with hazards
2. Transient conditions → Temporary conditions detected as hazards
3. Camera coverage → Hazards in blind spots not detected

Cross-References

• safety-PPE-detection/SKILL.md — PPE compliance detection
• safety-alert-system/SKILL.md — Alert generation