Development of a markerless optical motion capture system for daily use of training in swimming

The main objective of this research was to develop a markerless optical motion capture system that can be used for daily use in swimming training. The butterfly stroke was targeted since it is considered bilaterally symmetric in motion. The system consisted of a segmentation process to obtain the participant’s silhouettes and a matching process to estimate the pose of the participant. A variable thresholding method was used to extract the silhouettes to solve non-uniform illumination in the recorded swimming video. Prior to the matching process, the human body was modeled as a series of nine segments to help the matching process. The model was then mapped so that it aligned with the silhouettes, which were investigated by similarity of intensity value. To minimize the degree of freedom in image matching, the available joint motion in the swimming human simulation model was used as a priori information for kinematics data of the swimming motion. As a result, the rotation angle’s correlation coefficients between the references and result of the matching process were around 0.95 for trunk, thigh, shank, upper arm, forearm, hand and 0.78 for head, hip and foot. The rotation angle and the velocity of the center of mass were put into the swimming human simulation model for a dynamics analysis. The simulation results show that the velocity obtained in the experiment corresponded to the fluid force exerted on the lower and upper limbs. Consequently, the proposed system of obtaining the joint motion of the butterfly stroke is suitable for daily training and coaching. © 2016, International Sports Engineering Association.