Related Experiment Video
Updated: May 17, 2025
09:41
Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping
Published on: April 21, 2023
1.4K
Performance Evaluation of Monocular Markerless Pose Estimation Systems for Industrial Exoskeletons
Soocheol Yoon1,2, Ya-Shian Li-Baboud1, Ann Virts1
1National Institute of Standards and Technology, Gaithersburg, MD 20899, USA.
Sensors (Basel, Switzerland)
|May 14, 2025
Summary
Markerless pose estimation systems show potential for analyzing human movement with industrial exoskeletons, but accuracy is affected by exoskeleton occlusion and viewpoint. Further development is needed for reliable human-robot interaction studies.
Area of Science:
- Biomechanics
- Robotics
- Human-Computer Interaction
Background:
- Industrial exoskeletons aim to reduce worker fatigue and injuries.
- Accurate human joint kinematics and human-robot alignment are crucial for understanding exoskeleton effects.
- Markerless pose estimation using RGB and depth cameras is emerging for human motion analysis.
Purpose of the Study:
- To evaluate the performance of monocular markerless pose estimation systems for human skeletal joint estimation while wearing industrial exoskeletons.
- To compare markerless systems against an optical tracking system (OTS) as a reference.
- To identify key factors influencing pose estimation accuracy in exoskeleton applications.
Main Methods:
- Evaluated two markerless pose estimation systems (RGB and depth) from ten viewpoints for one subject in 14 industrial poses.
- Repeated experiments with three different types of industrial exoskeletons.
- Utilized an optical tracking system (OTS) as the ground truth reference.
- Proposed a tracking quality ratio metric for assessing skeletal joint estimation acceptability.
Main Results:
- Image acceptance rates were 56% for RGB, 22% for depth, and 78% for OTS.
- Major sources of error included exoskeleton occlusions, specific industrial poses, and camera viewpoints.
- OTS performance degraded due to marker occlusion or shifting by the exoskeleton.
Conclusions:
- Markerless pose estimation systems offer a promising, non-invasive approach for exoskeleton research.
- Exoskeleton-induced occlusions and specific industrial poses significantly challenge current markerless systems.
- Further research is needed to improve the robustness and accuracy of markerless systems for reliable human kinematics analysis in exoskeleton studies.
