Automatic assessment of posture deviations in assembly tasks

The aim of this study is to investigate the development and evaluation a computer vision-based framework aid automatic assessment posture deviations in assembly tasks realistic work environments. A deviation refers time-varying working performed by worker, that deviates from ergonomically safe body postures expected context particular known impose increased physical strain. estimation their occurrences can serve as indicators, risk factors, for ergonomics towards prevention strain the-long-term work-related musculo-skeletal disorders (WMSD). Using visual information acquired camera sensors, our goal estimate full motion line worker 3D space, unobtrusively, perform classification four types deviations, also noted sub-optimal were employed MURI analysis tool. We formulate learning-based action task using Deep Graph-based Neural Networks differential temporal alignment cost measure type level observed during activities. To evaluate efficiency proposed approach, new video dataset was captured sustAGE project, demonstrate two different workers car door actions simulated production an actual workplace. Rich annotation data provided experts manufacturing ergonomics. Both quantitative qualitative provide evidence its reliability supporting ergonomic preventive WMSD real