A general marker-less motion capture approach for the Microsoft Kinect Sensor v2 using subject-specific articulated models within an iterative closest point algorithm for improved anatomical accuracy.

Gammelgaard, Mikkel Svindt

4. semester

Sports Technology (cand.tech.), Master

2016

2016-06-03

13 pages

Joint kinematics obtained by the Microsoft Kinect Sensors are measured based on a simplification of the human anatomy, which don’t include anatomically accurate estimations of the joint centers. Consequently the measured joint kinematics are not directly compatible with the anatomically accuracy of musculoskeletal models. Accordingly, the purpose of the present study was to propose a new marker-less tracking approach using the Microsoft Kinect v2 to obtain the 3D movement of a single leg. A subject specific articulated model of the right leg with anatomically correct joint centers was generated using the Microsoft Kinect v2 to scan the anatomical geometry of the segments. The poses of the detailed articulated model was estimated relative to a dynamic depth recording of a 90 degree hip flexion movement, using an iterative closest point algorithm to which a constrained optimization problem was defined. The proposed approach was compared to a marker-based system. Hip flexion/extension, abduction/adduction, external/internal, knee flexion/extension and ankle plantar/dorsi flexion were measured for one subject. The flexion/extension and internal/external rotation for the hip and flexion/extension for the knee showed good agreement with the marker-based results. The hip abduction/adduction and ankle flexion/extension showed poor results compared to the marker-based system.

Marker-less motion capture ; articulated model ; joint kinematics ; pose estimation ; kinematic analysis ; Microsoft Kinect ; iterative closest point

Download
View record in AAU Student Projects