Development and Validation of a Computer Vision Based System for Ergonomic Posture Evaluation Through Workplace Video Analysis

Work-related musculoskeletal disorders (WMSDs) often result from repetitive inappropriate postures and movements. The study develops a video-based work pose entry system for ergonomic postural assessment to evaluate postures based on the RULA and REBA assessment tool. The proposed work pose entry system generates RULA/REBA scores with little input required from the investigator and the Euler angle of each relevant joint, which is automatically computed by computer vision algorithms. Using the YOLOv3 algorithm for human tracking and SPIN for 3D pose estimation, this system processes 2D video inputs to generate scores with corresponding investigation levels. Traditional assessments depend on evaluator expertise, which is time-consuming, costly, and error-prone. To improve consistency, we conducted a study with 20 evaluators classified as experienced or novice based on their ergonomics knowledge. The proposed video-based work poses entry system for ergonomic postural assessment offers a reliable and efficient method for evaluating working postures. The reliability of the proposed tool is validated by comparing manual evaluation results.