Abstract
Background: Marker based motion capture (MBS) is the preferred method for 3D motion analysis in sports injury prevention. There is a continuous search for more efficient and low-cost methods that can produce equally reliable measures. Markerless motion capture systems (MLS), like the Kinect v1 and v2 are cheap commercial 3D sensors that have implications for injury risk research. The purpose of this thesis is to validate a markerless 3D motion capture system (MLS) against a marker-based motion capture system (MBS). Objectives: To evaluate the accuracy and tracking ability of MLS, Kinect v1 and v2 To investigate the side tracking ability of MLS To investigate the inter-trial reliability of MLS for each task To assess methods for synchronizing the data from the MBS and MLS in time and space To consider use of MLS in a clinical setting for knee motion assessment Material and Methods: During two screening sessions with 12 (3) Norwegian elite female basketball players performed three tasks; a vertical drop jump, two-legged squat (front/side) and single-leg squat (left/right). The MLS (Kinect v1/v2) and MBS (Qualisys) simultaneously recorded the respective testing sessions in a 3D motion analysis lab. Kinematic measures were calculated and compared in a statistical analysis. Results: Kinect v1 and 2 had very good correlation scores for mean PF during the two-legged squat, indicating a significant relationship between MLS and MBS. Almost all VDJ variables were not significantly different from the MBS (twotailed paired samples t-test, p < 0.05). KASR was poorly correlated for Kinect v1, except for mean PF during the squat(SRCC: 0.93) The two-legged squat scored highest overall in inter-trial reliability for all variables in session 1 and 2 (ICC: 0.81 - 0.99). FPPA was estimated more accurately by Kinect v1 than v2. Conclusion: Visual inspection of curves (PF and FPPA) for VDJ and squat, confirmed a real relationship between MLS and MBS variables. Adjustments to screening protocol will most likely improve the accuracy of the skeletal tracking (front and side) for Kinect v2 MLS. A simple method for estimating a common start-point in time for MLS and MBS, was found in session 1. Synchronization in space for kinematic measures require defining local coordinate systems for each body segment in MLS skeleton. Implementation of MLS in a clinical setting require further testing and verification of its accuracy in motion tracking.