Feasibility of a single camera markerless motion capture system to accurately measure flexion-extension

INTRODUCTION Traditional motion capture techniques use markers to measure movement. However, markers are prone to soft tissue artifact and experimental error [1] as well as possible impedance to natural motion. Another significant challenge is the ability to use optical motion capture systems in a variety of natural setting such as a patient’s home, on the sports field, or in public. This has led to the development of markerless motion capture [2]. The accuracy of marker-based systems has been analyzed using custom built devices [3], in vivo imaging [4], external fixators [5], and electrogoniometers [6]. By comparison, the accuracy of markerless motion capture techniques in calculating joint angles has not been studied as extensively. Joint kinematics calculated from marker-based and markerless motion capture have been compared [7], but this only provides a measure of relative accuracy. The absolute accuracy of markerless motion capture has been analyzed using a virtual walking model, but this excludes errors due to camera calibration and foreground/background separation [8]. Also these systems used an array of cameras and were not systems such as the Microsoft Kinect which relies on just two cameras: a depth map and a color camera. Therefore, the goal of this study was to quantify the accuracy of marker-based and markerless motion capture systems in the calculation of joint angles during a static posture and to compare the differences in joint angles during motion using a simulated leg in an experimental setting.