Ergovision: Computer Vision-Based Ergonomic Risk Assessment for Industrial Work in Resource-Constrained Settings

Work-related musculoskeletal disorders (WMSDs) impose significant health and economic burdens globally, with disproportionate impact in Sub-Saharan Africa where over 600 million people lack electricity access and industrial expansion is accelerating[1]. Traditional ergonomic assessment methods like the Rapid Entire Body Assessment (REBA) require trained experts and manual observation, limiting scalability in resource-constrained environments[2]. This paper presents Ergovision, an open-source computer vision system that automates REBA-inspired ergonomic risk assessment from monocular RGB video using MediaPipe Pose for 3D keypoint extraction. The system computes neck, trunk, and lower-limb joint angles, maps them to risk scores (negligible to very high), performs temporal analysis of posture changes, and generates practitioner-ready HTML reports. Validation against manual assessment by certified ergonomists across 36 industrial tasks showed 78% agreement (Cohen's κ = 0.71), comparable to inter-rater reliability of manual methods. Designed for deployment on commodity hardware (Intel Core i5, 8GB RAM, standard webcam), Ergovision reduces assessment costs by over 80% while enabling continuous monitoring. We describe the system architecture, validation results, and deployment considerations for Africa's clean-energy and industrial sectors.