The throwing distance in the shot put is determined by release velocity, angle of release, height of release. Those kinematic variables have traditionally been collected using a gold standard system for motion capture which requires a considerable amount of time, money, and professional knowledge. This challenging setup can be overcome by using OpenPose which is the most validated pose estimation system among other systems. However, most validation studies have not been focusing on athletic and sports movements including shotput throwing. Therefore, this study aims to validate the OpenPose measurement for shotput by using Optitrak. A former shot-put thrower performed 32 throws (4 different throws × 8 attempts in each) which were recorded by 12 Optitrak and 4 CASIO cameras. We extracted 12 joint positions including the wrist, elbow, shoulder, hip, knee, and ankle on the right and left side by using the OpenPose and reconstructed 3D data by using direct linear transformation (DLT). We also collected the same 12 joint points from the Optitrak and calculated MPJPE (Mean Per Joint Position Error) to evaluate the accuracy of the OenPose. The result showed that the right wrist was the most erroneous joint among other joints. Also, more errors were observed in the Spin and Half spin throws compared to the other throws. From the result, the OpenPose can be used for the performance analysis with caution, but further research needs to be conducted to investigate the use case of the OpenPose in athletic performance analysis.
«
The throwing distance in the shot put is determined by release velocity, angle of release, height of release. Those kinematic variables have traditionally been collected using a gold standard system for motion capture which requires a considerable amount of time, money, and professional knowledge. This challenging setup can be overcome by using OpenPose which is the most validated pose estimation system among other systems. However, most validation studies have not been focusing on athletic and...
»