فهرست نشریات علمی دسترسی آزاد ایران

Ensuring worker safety is a critical priority in the construction industry. While Building Information Modeling (BIM) has advanced project management and visualization, its role in spatially-aware safety analysis, particularly for Personal Protective Equipment (PPE) violations, remains limited. Existing approaches often fail to capture insights into high-risk zones. This study proposes a framework for spatial risk categorization within BIM models, based on the frequency of PPE violations. Using a computer vision approach with the You Only Look Once (YOLO) model, PPE infractions involving hard hats and safety vests are automatically detected from site imagery and aggregated per spatial element. These violations are linked to the Revit model, incorporating camera positions to visualize high-risk zones as color gradients within the model. The evaluation demonstrates strong detection accuracy, with mean average precision (mAP) values of 0.823 for “Person,” 0.819 for “Hat,” and 0.567 for “Vest,” yielding an overall mAP of 0.746. By highlighting spatial zones with elevated risk, the framework supports targeted deployment of safety measures where needed. It also enables tailored training programs for subcontractor crews in these zones, ensuring context-specific and effective safety management. This innovative approach introduces a new data layer to coordination models, derived from real-world safety performance, enabling stakeholders to spatially identify areas with high PPE non-compliance. This enables proactive monitoring and allows for the timely deployment of safety measures. Informed decision-making is thereby supported, leading to more effective safety interventions and a reduction in on-site hazards. The study's finding aligns with the perspectives of safety experts, validating a practical approach to hazard reduction.