System and modelling errors in motion analysis: Implications for the measurement of the elbow angle in cricket bowling

Abstract

The system and modelling errors of two fundamentally different motion capture systems (opto-reflective vs. video-based) were tested under various conditions, to determine their ability to accurately measure flexion–extension of the elbow angle in cricket bowling. A mechanical arm was used for all testing, that enabled known elbow flexion–extension and abduction (“carry”) angles to be manually set. The root mean squared (RMS) error of 0.6° for the opto-reflective system (Vicon-612) was more accurate in reconstructing a known angle than the video-based system (Peak Motus) (RMS error 2.3°) in the laboratory, when the same mathematical procedure (model) was applied to calculate the elbow flexion–extension angle. When different models were applied to the raw marker trajectories collected using the video-based system, RMS was lowest for the external marker segmental cluster models (2.3°) compared with 9.4° for the vector and 4.5° for the projected vector approaches, where joint centres were visually approximated. Real world, field-based comparisons using the video-based system showed that occluding the arm and therefore the shoulder, elbow and wrist joint centre locations by placing a shirt on the arm, increased RMS error for both vector (7.8°–9.0°) and projected vector (4.3°–5.1°) modelling approaches.