MarkerLess Motion Capture

MarkerLess Motion Capture: ML-MoCap, a low-cost modular multi-camera setup

Geelen, J.E. (TU Delft Biomechatronics & Human-Machine Control)
Branco, Mariana P. (University Medical Center Utrecht)
Ramsey, Nick F. (University Medical Center Utrecht)
van der Helm, F.C.T. (TU Delft Biomechatronics & Human-Machine Control)
Mugge, W. (TU Delft Biomechatronics & Human-Machine Control)
Schouten, A.C. (TU Delft Biomechatronics & Human-Machine Control)

2021

Motion capture systems are extensively used to track human movement to study healthy and pathological movements, allowing for objective diagnosis and effective therapy of conditions that affect our motor system. Current motion capture systems typically require marker placements which is cumbersome and can lead to contrived movements.Here, we describe and evaluate our developed markerless and modular multi-camera motion capture system to record human movements in 3D. The system consists of several interconnected single-board microcomputers, each coupled to a camera (i.e., the camera modules), and one additional microcomputer, which acts as the controller. The system allows for integration with upcoming machine-learning techniques, such as DeepLabCut and AniPose. These tools convert the video frames into virtual marker trajectories and provide input for further biomechanical analysis.The system obtains a frame rate of 40 Hz with a sub-millisecond synchronization between the camera modules. We evaluated the system by recording index finger movement using six camera modules. The recordings were converted via trajectories of the bony segments into finger joint angles. The retrieved finger joint angles were compared to a marker-based system resulting in a root-mean-square error of 7.5 degrees difference for a full range metacarpophalangeal joint motion.Our system allows for out-of-the-lab motion capture studies while eliminating the need for reflective markers. The setup is modular by design, enabling various configurations for both coarse and fine movement studies, allowing for machine learning integration to automatically label the data. Although we compared our system for a small movement, this method can also be extended to full-body experiments in larger volumes.

http://resolver.tudelft.nl/uuid:62c03f69-82fe-4c34-8c17-bdf3ee32421c

https://doi.org/10.1109/EMBC46164.2021.9629749

IEEE

978-1-7281-1179-7

Proceedings of the 43rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2021

43rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2021, 2021-11-01 → 2021-11-05, Virtual, Online, Mexico

Institutional Repository

conference paper

© 2021 J.E. Geelen, Mariana P. Branco, Nick F. Ramsey, F.C.T. van der Helm, W. Mugge, A.C. Schouten