A Fundamental Study on Development of a Workload Evaluation System Based on NLE Using Two-Dimensional Skeleton Data

In many production sites, health issues such as lower back pain are often caused by high-workload tasks. Therefore, to establish safe and worker-friendly environments by quantitatively assessing and managing the workload is one of a management challenges. However, it is both technically and economically difficult to consistently quantify and evaluate the workload in actual production sites, and the evaluation remains limited. To solve these problems, this study focused on material-handling work (loading and unloading work) and developed a system that automatically evaluates the workload using work videos taken with a camera. Specifically, a methodology is devised to apply NIOSH Lifting Equation using two-dimensional skeleton date from MediaPipe (skeleton estimation framework) to determine origin and destination of material-handling work and to estimate actual distance and angle data. Then, the accuracy and usefulness of the system are verified by comparing it with data from optical motion capture. As results of the laboratory experiment, it is found that determining the accuracy of the origin and destination was 94.74%. In addition, although the estimation accuracy of the actual distance data showed a small error, a correlation with the actual measured values (calculated by the optical motion capture) were confirmed. However, the estimation accuracy of the angle was low, and no correlation with the actual value was confirmed. From the above, the workload evaluation system based on NLE can quantify the workload under limited conditions. In the future, the accuracy of actual distance and angle estimation will be improved, and the system will be applied to actual production sites to evaluate its practicality.