Development of an ergonomic risk assessment system for repetitive tasks by skeleton recognition based on MediaPipe

In product manufacturing, certain repetitive tasks may be altered due to the combination of work procedures, leading to ergonomic injuries among workers. To reduce the occurrence of occupational accidents and enhance the efficiency of ergonomic evaluations, this study develops an ergonomic hazard assessment system based on the MediaPipe skeleton recognition framework, combined with the ResNet-18 model. The system performs posture recognition on images or videos of workers, automatically detecting their movements and marking potential poor postures. The image recognition accuracy of the system exceeds 80%. The findings from this study can assist workers in preventing work-related musculoskeletal disorders (WMSDs), enhancing productivity, and mitigating ergonomic risks.