UAV-DL-Based Computer Vision Framework for Construction Worker Pose Estimation

Traditional ergonomic assessments in construction are limited by subjectivity, intermittent monitoring, and scalability challenges. To address these issues, this study presents a framework integrating unmanned aerial vehicles (UAVs) and deep learning-based computer vision for automated ergonomic risk assessment of construction workers. The framework utilizes UAVs to capture aerial footage of workers, enabling continuous monitoring across large and complex sites. Leveraging MediaPipe for pose estimation, the framework detects key body landmarks and calculates joint angles, providing an objective representation of worker postures. These joint angles are then mapped to the Rapid Entire Body Assessment (REBA) scoring system through custom algorithms, which translate postural data into standardized ergonomic risk scores. By integrating pose estimation with REBA scoring, this framework offers an automated, data-driven approach to identify and categorize ergonomic risks, facilitating proactive interventions to reduce musculoskeletal disorder (MSD) risks on construction sites. This study demonstrates the potential of UAVs and computer vision to enhance construction safety through scalable, objective, and continuous ergonomic assessment.