Study at a glance
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 v2025.1.6, as stated in the preprint.
Study Overview
Design
A within-subject, repeated-measures study comparing lower-limb gait kinematics from identical walking trials recorded simultaneously across four camera height groups and three capture-volume widths, using Theia3D markerless motion capture.
Participants
Nine healthy adults (5 men, 4 women), all over 18 years old, with no history of gait impairments.
Camera Layouts
Thirty-two video cameras were arranged elliptically around a central walkway in four height groups (0.85 m, 1.65 m, 2.45 m, and 3.30 m; eight cameras per group) and tested across three capture-volume widths (6.9 m, 5.5 m, and 3.3 m), with capture volume length held constant at 13.8 meters.
Data Collection
Participants walked overground at a self-selected comfortable pace along a 10-meter walkway, walking back and forth twice at each capture width within a single testing day, wearing their own casual clothing.
Analysis
Three-dimensional lower-limb kinematics were calculated independently for each camera grouping. Linear mixed-effects models assessed differences in joint-angle waveforms by camera height group, with follow-up pairwise comparisons at seven relevant points in the gait cycle. Reliability across the four simultaneously recorded camera setups was quantified using the intraclass correlation coefficient (ICC) and standard error of measurement (SEM).
Key Findings
- Significant kinematic differences appeared across every lower-limb joint and capture width tested, with the highest camera placement (3.30 m) driving 78% of significant pairwise comparisons and the lowest placement (0.85 m) driving 42%, while the two intermediate heights (1.65 m and 2.45 m) showed no significant differences from each other at any of the seven discrete gait events examined.
- Overall agreement remained favorable despite these statistically detectable differences. Integrated reliability coefficients averaged 0.89 across all conditions, and integrated measurement error averaged 0.56 degrees, with pointwise error staying below 2 degrees throughout the gait cycle in every condition tested.
- The narrowest capture width consistently showed the lowest agreement, with reduced reliability most apparent during the swing phase of the gait cycle, a pattern the authors attribute to steeper camera angles and less camera overlap at narrower widths, which produce less favorable triangulation geometry and more self-occlusion.
- Joint-specific patterns showed where occlusion mattered most. At the ankle, frontal-plane variability was greatest and increased as capture width narrowed, with the largest error occurring during pre-swing, when the ankle is increasingly hidden from view by the knee and thigh from elevated, anterior camera angles; even there, measurement error stayed below 1 degree.
- Mid-range camera heights consistently produced the greatest agreement across all conditions tested, while both the highest and lowest camera placements introduced the largest deviations, supporting a specific, evidence-based recommendation for camera height standardization.
What This Means for Movement Science Researchers and Lab Methodologists
For labs setting up or expanding a markerless capture system, this study offers a concrete, actionable recommendation: standardizing camera height within the 1.65- to 2.45-meter range produced the most consistent lower-limb kinematics across every condition tested, which could meaningfully improve comparability of data collected across different rooms, sites, or multicenter studies. At the same time, the authors are clear that camera placement was a modest source of variation relative to other previously documented sources of error in markerless motion capture, so applications especially sensitive to subtle between-group differences or small longitudinal changes may benefit most from stricter placement standardization, while less precision-sensitive applications may have more flexibility in how cameras are arranged.
The Role of Theia3D in Movement Analysis
Theia3D processed all 32 camera configurations independently in this study, generating lower-limb kinematics from each height-and-width combination so that camera placement, held separate from any change in the walking task itself, could be isolated as a variable and evaluated on its own. This approach let the researchers attribute measured kinematic differences specifically to camera geometry rather than to differences in participant movement between trials.
Study Limitations and Future Directions
The authors note that no marker-based reference was used in this study, since marker-based systems carry their own well-documented sources of error and don't represent true, error-free skeletal motion; as a result, this study evaluates agreement between different markerless camera configurations rather than absolute measurement accuracy against a ground truth. Some of the reduced agreement observed during the swing phase may reflect sensitivity to small timing shifts introduced during gait-cycle normalization rather than true camera-configuration differences. Different walking trials were performed at each capture-width condition rather than being collected simultaneously, so some of the measured variability may reflect ordinary trial-to-trial differences rather than camera placement itself. Narrower capture-width conditions also qualified fewer strides and participants for analysis, which may have produced less stable reliability estimates independent of any true change in kinematic agreement. The authors frame mid-range camera height standardization as their core practical recommendation, while noting that camera placement should be considered one methodological factor among several, rather than the dominant source of variation, provided extreme configurations are avoided.
Citation: Hadwen, A. J., Outerleys, J. B., & Deluzio, K. J. (2026). Camera placement influences markerless lower-limb kinematics. SSRN. https://papers.ssrn.com/abstract=7238027
Link: Read the full preprint
Where You Put the Cameras Matters, Within Reason
This study offers labs a specific, evidence-based answer to a question that's usually settled by available space alone: camera height does measurably affect markerless kinematics, but staying within a mid-range band keeps that effect small. If your work involves setting up or standardizing a markerless capture system across multiple rooms or sites, talk to our team about how camera configuration could affect your own data.