Implementation of Computer Vision Based Algorithms for Bio-Mechanical Risk Estimation in Upper Limbs During Activities With Repetitive Movements

Bio-mechanical risk, which implies the possibility of facing adverse events during work activities, is intrinsically linked to the occurrence of musculoskeletal disorders and affects various occupations. This category of impairment, predominantly associated with occupational diseases, has a variable impact depending on the specific characteristics of the performed task. Factors such as repetitive movements, forced postures, and handling of loads, among others, can trigger this risk. In order to prevent and control this risk, it is essential to carry out diagnoses and evaluations using validated methods or instruments. The acquisition of data through these instruments requires techniques that minimize subjectivity and reduce the time needed for data capture. In this context, this document describes the preliminary results of research that implements computer vision-based algorithms to assess bio-mechanical risk associated with repetitive movements in the upper limbs. This evaluation was conducted on workers in the pharmaceutical industry, specifically in production and conditioning activities. The risk assessment methodology used as a basis was the OCRA checklist, and computational algorithms based on the “You Only Look Once” YOLO and YOLO-Pose models were implemented, obtaining promising results from the implementation of a semi-automated bio-mechanical risk measurement model.