Presentation
Human Bounding Boxes for Efficient Computer Vision Estimation of the Revised NIOSH Lifting Equation
SessionPoster Session 1
DescriptionManual lifting assessment relies on estimation of the hand location, particularly for the horizontal (H) and vertical (V) distances in the Revised NIOSH Lifting Equation (RNLE). Previous computer vision-based approaches depended on complex 2D/3D pose estimation. While these methods are effective, they are computationally intensive, and they are sensitive to occlusions and visual interference. This study proposes a simple, alternative: using the dimensions of a bounding box (BBX) in the sagittal plane from RGB video, normalized by the person’s standing height, to estimate H and V, and consequently, the Recommended Weight Limit (RWL). To test this hypothesis, we utilized laboratory data comprising synchronized video and motion capture (MoCap) recordings from 10 participants performing 36 trials, lifting a tote pan from a platform at adjustable heights based on individual vertical landmarks and three horizontal distances (25.4 cm, 45.7 cm, 71.1 cm). Linear regression analysis on our experimental data supports the existence of the proposed relationship, confirming that normalized BBX dimensions are predictors of the normalized hand location. This study demonstrates that normalized BBX dimensions can predict hand location (H, V), achieving sufficient accuracy while requiring significantly less computational complexity than full body pose estimation.
Event Type
Poster
TimeTuesday, October 14th5:30pm - 6:30pm CDT
LocationRiverside East
Similar Presentations