Related Experiment Video
Updated: May 26, 2026
Movement Retraining using Real-time Feedback of Performance
Published on: January 17, 2013
Optimizing camera placement in multi-view markerless motion capture: a validation study against gold-standard
Ali Sahraneshin Samani1,2, Mohsen Razeghi3, Mohammad Abdoli-Eramaki4
1Student Research Committee, Shiraz University of Medical Sciences, Shiraz, Iran.
Objective:
This study aimed to determine the optimal camera configuration for a markerless motion capture system, which may offer a lower-cost alternative to laboratory based systems. It validated the system's performance against a gold standard during both static postures and a dynamic lifting task, addressing a key gap regarding the impact of camera arrangement on measurement validity. Six unique stereo camera pairs were calibrated for 3D reconstruction using MediaPipe. Joint angles were computed, and agreement was assessed using Mean Absolute Error (MAE), Intraclass Correlation Coefficient (ICC), and Bland-Altman analysis.
Results:
System performance was highly sensitive to camera geometry. The optimal configuration (approximately 45 degrees separation) demonstrated excellent agreement for both static (ICC = 0.851, MAE = 9.28 degrees) and dynamic (ICC = 0.823, MAE = 12.92 degrees) tasks. Major peripheral joints (hips, knees, shoulders) showed good to excellent reliability (ICC = 0.73-0.92), while neck showed only moderate agreement (ICC = 0.56-0.66). Configurations with very wide baselines (greater than 90 degrees) performed poorly (ICC < 0.40). A two-camera setup with a narrow baseline can yield clinically useful joint angle estimates for major joints, providing practical guidance for affordable motion-capture deployment in biomechanics, ergonomics, and rehabilitation.
