Computer vision-assisted multi-objective spatial optimization of fall protection systems in construction: Integrating hazard zone modeling and posture detection

Falls from height (FFH) remain the leading cause of fatalities in construction, highlighting persistent challenges in personal fall protection system (PFPS) planning. Despite regulations, anchorage placements still rely on subjective judgment and static layouts, limiting adaptability to complex site risks. This study develops a computer vision-assisted optimization framework integrating hazard zone modeling and worker posture detection. Vision-based posture data and hazard zone models construct spatial risk fields, providing a basis for anchorage planning. A multi-objective model is formulated to enhance safety performance and reduce swing fall risk, while a simulation module based on genetic algorithms computes Pareto-optimal layouts. Computer vision posture detection is embedded into the iterative module, enabling adaptive adjustments to deviations between planned and observed postures. A high-rise piping construction case study demonstrates the framework’s effectiveness in producing safety-resilient and efficient anchorage plans. The proposed method advances PFPS toward intelligent and data-driven safety management. • A computer vision-assisted framework integrates 3D site geometry, hazard zone modeling, and posture detection. • Multi-objective spatial optimization balances safety performance, task coverage, and swing-fall risk. • A genetic algorithm simulation module (SR-MOGA) generates Pareto-optimal anchorage layouts for safety-productivity trade-offs. • Real-time posture detection feedback enables adaptive reconfiguration of fall protection system planning.