Development of a Semi-Automatic Rapid Entire Body Assessment Model using the Open Pose and a Single Working Image

Abstract

Ergonomic risk assessment of work-related musculoskeletal disorders is crucial for workers’ safety and health. The purpose of this study was to develop a semi-automatic model of Rapid Entire Body Assessment (REBA), which is one of the representative ergonomic risk assessment methods. This study adopted Tensorflow implementation of the Open Pose that can detect the three-dimensional locations for 17 key points of a human body using a single image. A graphical user interface was designed in Matlab to visualize working posture with the key points and ask a user additional inputs (e.g., weight of working object, coupling quality of a handle) that can’t be detected from an image. Five actual working images were assessed by the proposed system and ergonomic experts to compare the resulting agreement. The assessment results revealed that similar risk scores were observed in the upper arm, lower arm, and wrist; but the proposed system slightly underrated for the trunk, neck and legs. The proposed system can efficiently help occupational safety and health (OSH) practitioners in assessing ergonomic risk of works using REBA.