Investigating the effect of parallax on upper limb posture analysis

Research Trainee: Michael Lau, PhD Student, Dept. of Industrial and Operations Engineering at the University of Michigan

Faculty Researcher: Thomas J. Armstrong, PhD, Professor of Industrial and Operations Engineering and biomedical Engineering at the University of Michigan

Observational methods are widely used to estimate work postures for application of job analysis tools used to evaluate stresses associated with hand and wrist musculoskeletal disorders. Often, observations for these methods, which include RULA, OWAS, VIDAR, Strain Index, ACGIH TLV for HAL, REBA, and GM-RFCII, are made retrospectively from video recordings. Available studies show that these estimates may be subject to significant error (Lowe, 2004). The aim of this study is to investigate factors that affect the accuracy and precision of wrist posture estimates based on observations from video recordings. Specific factors include parallax, image size and hand postures.

The effect of parallax on the angle between two intersecting lines can be easily computed and show that the perceived angle changes as a tangent function of the viewing angle. Unless the camera or observer is perfectly aligned with the axis of rotation, there will always be parallax. As a practical matter, it is almost impossible to not have some parallax error for wrist postures; an aspect seldom accounted for in existing literature.

The hand is a solid object, not simply two intersecting lines. The surfaces of the hand have shape, texture and color that provide visual queues that can help observers compensate for parallax. We believe that while these cues assist in posture estimation, they are insufficient to eliminate the effects of parallax.

This project has two aims. First, we will systematically investigate the effect of camera angle, image size and hand posture on observer estimates of static wrist postures. Second, we will determine the effect recording angle has on observer identification of peak and awkward postures in repetitive dynamic wrist motions. We will attempt to relate the findings to relevant risk assessment tools. This project will lay important foundations for research that examines the way ergonomic data are collected; the long term goal being to create a standardized approach to the recording and use of video data. Results can be used to develop preliminary correction factors to reduce observational error introduced by non-ideal video recording angles. Results may also prompt ergonomics researchers to begin reporting video capture methods so that results may be interpreted more carefully. By reducing exposure assessment error, the results will lead to improved exposure-response models and will help practitioners make fewer errors identifying workplace stresses and problematic jobs.


Research trainee’s current position:
Michael Lau completed his PhD in 2010 and is currently the Senior Human Factors Engineer at Insight Product Development.