Validation, Methods & System Performance
Sony RX0-II

The Effect of Camera Configuration on Multi-Camera Markerless Motion Capture for Biomechanical Analysis

Journal of Biomechanics
•
2026
•
Original research
Theia3D version:
v2024.1.17
More Articles

Study at a glance

In 10 healthy adults performing walking, running, and countermovement jumping, systematically reducing a Theia3D camera array from 18 to 12 to 6 cameras produced only small mean kinematic differences, less than 4 degrees across all tasks and planes, provided the capture volume and camera geometry preserved adequate segment visibility.

Summary

Disclaimer: The following summary details independent academic research. Theia3D is intended for research, sports performance and educational use only. It is not cleared as a medical device and is not intended to be used for the diagnosis, treatment, mitigation or prevention of any disease, injury or medical condition.

Editor's note: Theia3D version v2024.1.17, as stated in the paper.

Study Overview

Design

A within-subject, repeated-measures study examining how reducing camera number (18, 12, or 6 cameras), and reducing capture volume size (10 × 8 m or 6 × 4 m), affects Theia3D-derived joint kinematics during walking, running, and countermovement jumping, all generated by subsetting a single, simultaneously recorded 18-camera dataset for each trial.

Participants

Ten healthy adults (7 men, 3 women; mean age 31.0 years; mean height 1.8 m; mean mass 73.3 kg), excluded if they had a neuromuscular, musculoskeletal, or significant visual condition, or a cognitive or behavioural condition that would prevent them from following study instructions.

Data Collection

Eighteen synchronized Sony RX0M2G cameras (120 Hz, 1920×1080 resolution) were mounted on tripods at a height of 1.6 m, spaced at equal 20-degree intervals around the capture volume. Participants completed five trials each of walking and running at self-selected speeds, and five countermovement jump trials with hands on hips, across both a 10 × 8 m and a 6 × 4 m capture volume. The full 18-camera dataset was processed in Theia3D (v2024.1.17), and reduced 12- and 6-camera configurations were generated afterward by systematically selecting subsets of the same video data, so that any differences reflected camera configuration rather than differences in the underlying movement or calibration approach.

Analysis

Joint angles at the ankle, knee, hip, pelvis, and trunk were compared across camera configurations in the sagittal, frontal, and transverse planes using one-dimensional statistical parametric mapping (1D-SPM). For the larger 10 × 8 m volume, a repeated-measures analysis of variance compared all three configurations (18, 12, and 6 cameras), followed by post-hoc pairwise tests; for the smaller 6 × 4 m volume, a paired t-test compared the 18- and 12-camera configurations only, since the 6-camera configuration could not track all segments throughout every movement in that smaller volume.

Key Findings

  • Reducing camera number from 18 to 12 to 6 produced only small mean kinematic differences in the larger 10 × 8 m capture volume. Mean differences across all camera configurations, joints, and planes stayed below 2 degrees for walking, below 4 degrees for running, and below 3 degrees for countermovement jumping.
  • The largest differences in the 10 × 8 m volume consistently involved the 18-versus-12-camera and 12-versus-6-camera comparisons, while the smallest differences occurred at the trunk across every movement tested. During walking, the largest differences occurred at the pelvis (up to 1.7 degrees in the sagittal plane); during running, the largest differences occurred at the hip (up to 1.7 degrees across all three planes); during countermovement jumping, the largest differences occurred at the knee and ankle in the transverse plane (up to 2.3 and 2.6 degrees respectively).
  • In the smaller 6 × 4 m volume, the 18- and 12-camera configurations remained broadly comparable, but the 6-camera configuration failed outright, producing incomplete tracking and requiring camera repositioning to be usable. Where the 6-camera configuration could be analyzed in the larger volume, it remained viable; in the smaller volume, insufficient camera coverage led to tracking failures rather than a measurable kinematic difference.
  • Countermovement jumps performed centrally within the capture volume showed minimal differences between the 18- and 12-camera configurations in the smaller volume, with all mean differences below 1.3 degrees, while running trials that moved closer to the edges of the smaller volume showed larger differences, particularly at the ankle during the last quarter of the gait cycle (up to 4.1 degrees across all three planes).
  • The authors interpret camera geometry, specifically whether a given configuration maintains consistent visibility of each body segment throughout a movement, as the primary factor driving the differences observed, rather than camera count on its own. Differences between the 12-camera and 18-camera configurations were attributed in part to the 12-camera setup's more front-facing camera distribution, which changed the angular diversity of views contributing to 3D reconstruction compared with the evenly spaced 18- and 6-camera arrangements.

What This Means for Movement Science Researchers and Lab Methodologists

For teams weighing whether a smaller camera array or more constrained capture space could still support reliable Theia3D-derived kinematics, this study offers encouraging, practically specific guidance: reducing camera number from 18 down to 12, and in some cases to 6, produced only modest mean kinematic differences across walking, running, and jumping, as long as the resulting configuration maintained adequate visibility of each tracked body segment throughout the movement. The clearest caution is that camera count alone isn't the deciding factor, camera geometry and capture volume together determine whether a reduced configuration will actually work, and a configuration that performs well in a larger space can fail outright in a smaller one if segment visibility isn't preserved. For teams planning a leaner or more portable Theia3D setup, these findings suggest the design question to prioritize is maintaining consistent segment visibility across the movements of interest, rather than simply matching camera count to a published reference setup.

The Role of Theia3D in Movement Analysis

Theia3D was the markerless motion capture system evaluated throughout this study, processing the same simultaneously recorded video data under three systematically generated camera configurations and two capture volume sizes to isolate the effect of camera configuration itself on its kinematic output. This design allowed the researchers to directly attribute observed kinematic differences to changes in camera number and geometry, independent of any change in the underlying movement being captured, since each configuration was derived from the same original 18-camera recording.

Study Limitations and Future Directions

The authors note that no marker-based reference system was included in this study, so the findings reflect configuration-related differences between camera setups relative to each other, rather than the absolute accuracy of any configuration against a gold-standard measurement. A 6 Hz filtering cutoff was applied across all walking, running, and countermovement jump data, which may have attenuated some higher-frequency components during the more dynamic tasks, though this is expected to affect absolute kinematic magnitudes rather than the relative differences observed between configurations. A single calibration was applied across all camera configurations within each capture volume to isolate the effect of camera number specifically; when the researchers additionally tested independent calibrations for configurations with fewer cameras, residual calibration error increased slightly, suggesting this study's design may reflect a best-case calibration scenario relative to what independently calibrated, reduced-camera setups could show in practice. All camera configurations tested were based on subsets of the original 18-camera positioning; the authors note that alternative camera placement strategies designed specifically for a constrained space might perform differently than simply removing cameras from a layout built for a larger array. The authors identify examining optimized camera placement strategies for restricted environments, directly comparing capture volume sizes using configuration-specific calibrations, and extending this work to additional sport-specific tasks and participant populations as clear next steps.

Citation: Walker, R., Robinson, M. A., Outerleys, J., McKenzie, L., Janecic, S., Fragoso Vargas, N., Sumner, S., Winterbone, N., Lee, Y.-H., Gibbon, K. C., & Foster, R. J. (2026). The effect of camera configuration on multi-camera markerless motion capture for biomechanical analysis. Journal of Biomechanics, 206, 113522. https://doi.org/10.1016/j.jbiomech.2026.113522

Link: Read the full study

Does Camera Count Really Determine Markerless Capture Quality?

This study shows that reducing a Theia3D camera array from 18 down to 12, and in some cases 6, cameras produced only small kinematic differences during walking, running, and jumping, so long as the configuration preserved adequate visibility of each tracked body segment. If your work involves designing a more portable or space-constrained markerless capture setup and you want to understand how camera configuration could affect your own data, talk to our team about optimizing a Theia3D setup for your environment.

Recent Posts