Presentation
A Pose Correction Approach to Enhance the Kinematics Accuracy of Markerless Motion Capture During Manual Lifting Tasks
SessionPoster Session 2
DescriptionOcclusion due to human-object interaction remains a significant challenge in markerless motion capture (MoCap) during manual lifting tasks, resulting in substantial errors in joint kinematics, especially in the upper extremities. To address this challenge, we presented a novel approach to extract the motion pattern of the lifted object (box) from videos and utilize it to reconstruct occlusion-induced missing wrist joint positions. In a laboratory study, eight healthy adults completed symmetric box lifting tasks, which were recorded using two synchronized smartphones (markerless MoCap) as well as an 8-camera optical motion capture system (marker-based MoCap). Smartphone videos were processed to extract joint kinematics with and without the proposed pose correction approach. The kinematic accuracy of the markerless system was evaluated by calculating the root mean square differences (RMSDs) between the markerless and marker-based systems. The results showed that the pose correction approach significantly reduced kinematics errors (i.e., RMSD) of the markerless system for elbow, shoulder, and lumbar flexion angles. The findings suggest that the proposed approach can reduce occlusion-related tracking errors and enhance the kinematics accuracy of markerless MoCap during manual lifting tasks.
Event Type
Poster
TimeWednesday, October 15th5:30pm - 6:30pm CDT
LocationRiverside East