Author(s): PASUNURU, S., PANCHAGNULA, M.V., Institution: INDIAN INSTITUTE OF TECHNOLOGY MADRAS, Country: INDIA, Abstract-ID: 2238

INTRODUCTION:
Spinning balls play a critical role in sports such as cricket, tennis, table tennis, and football. Among these, table tennis provides a distinctive testbed for studying ball aerodynamics, as drag and lift forces can exceed gravitational influence, producing pronounced drift during flight [1]. Spin also has a strong influence on post-impact behavior, significantly affecting bounce characteristics. Existing bounce models are typically tailored to specific interaction regimes and lack general applicability [2]. Understanding both flight and bounce mechanics requires precise measurement of position, velocity, and spin from experimental trajectories. Accurate kinematic extraction enables reliable trajectory prediction and provides key parameters immediately before and after impact, which are essential for analyzing bounce dynamics.

METHODS:
A high-speed stereo vision system was developed to capture the three-dimensional trajectories of sports balls, including the bounce phase. In this study, the system is applied to table tennis shots with pure topspin, backspin, sidespin, and their combinations. A custom computer vision pipeline processes synchronized footage from the two cameras to reconstruct the ball’s 3D position over time. Velocity vectors are computed at each timestep from the reconstructed positions. An inhouse mathematical framework is further implemented to estimate the instantaneous spin vector using information from either camera view.

RESULTS:
The position and spin extraction algorithms were validated against ground-truth references. The vision pipeline achieves sub-millimeter accuracy in reconstructing the ball’s position. Spin rate estimation errors were consistently below 5%.
CONCLUSION:
The developed pipeline provides a robust framework for extracting trajectory and spin parameters of spherical sports balls. The methodology is adaptable across multiple sports, and the resulting dataset of experimentally derived kinematics enables improved understanding and modeling of both aerodynamic behavior and bounce mechanics.
[1] Clanet 2015 [2] Cross 2002