Camera Placement Influences Markerless Lower-limb Kinematics

Markerless motion capture technology is widely used in clinical and research biomechanics, enabling collection of high-quality movement data outside traditional laboratory environments. However, the extent to which camera position affects kinematic outcomes has not been fully characterized. This study evaluated the effect of camera placement on lower-limb kinematics during overground walking. Nine healthy adults performed trials across three capture volume widths (6.9 m, 5.5 m, 3.3 m), with movement recorded using thirty-two synchronized video cameras separated into four distinct height groups (0.85 m, 1.65 m, 2.45 m, 3.30 m) and processed independently in Theia3D (v2025.1.6). Significant kinematic differences were observed across all lower-limb joints and capture widths, with the highest camera placement group (3.30 m) driving 78% of significant pairwise comparisons, while the two intermediate heights (1.65 m and 2.45 m) did not differ significantly at any discrete gait event. Despite these findings, overall agreement remained favorable: integrated intraclass correlation coefficient values averaged 0.89 ± 0.09 and integrated standard errors of measurement averaged 0.56° ± 0.13, with pointwise SEM remaining below 2° across all conditions. Agreement was lowest in the narrowest width condition, with reduced reliability most apparent during swing phase. These findings indicate that camera placement exerts a statistically detectable but practically modest influence on markerless gait kinematics. Mid-range camera heights consistently produced the greatest agreement, whereas extreme placements introduced the largest deviations. Standardization of cameras at mid-range heights (1.65 m – 2.45 m) is recommended for reproducible gait kinematics across different laboratories, environmental settings, and multicenter studies.