Determining the accuracy and concurrent validity of a computer vision system for running biomechanics analysis

Assessing running biomechanics is crucial for optimising performance and supporting rehabilitation in field-based team sports. While marker-based three-dimensional motion capture systems are considered the reference standard for assessing kinematics in controlled environments, their high cost and limited feasibility restrict on-field application. Recent advancements in computer vision offer more accessible, non-invasive alternatives. This study evaluated the accuracy and concurrent validity of VueMotion, a 2D camera-based computer vision system, for measuring sprinting kinematics. Twenty-five elite and sub-elite athletes completed four maximal 10-metre fly efforts (100 total trials), with their kinematics measured concurrently by VueMotion and a 3D motion capture system integrating infrared cameras and force plates. Root mean square errors for VueMotion peak velocity, ground contact time (GCT), and flight time were 0.19 m ⋅ s − 1 , 0.01 s, and 0.01 s, respectively. VueMotion mean biases (95% limits of agreement) were −0004 s (−0.014 to 0.005) for flight time, −0.016 m (−0.051 to 0.018) for step length, −0.09 m ⋅ s − 1 (−0.42 to 0.23) for peak velocity, and −0.14 m ⋅ s − 1 (−0.56 to 0.29) for instantaneous velocity. GCT and step frequency were 0.004 s (−0.004 to 0.013) and 0.031 Hz (−0.071 to 0.134), respectively. End users implementing the VueMotion system should interpret these findings relative to their assessment objectives, logistical resources, and operational constraints when applying the technology in field-based settings.