In order to develop a system capable of analysing table tennis matches in real time, we propose a method for estimating the ball dropped position. Previously, when the drop position of a hit ball was estimated using a regular commercially available camera, the error between the actual and estimated drop position was 140 mm on average with a maximum of ~300 mm. When predicting the drop position of a hit ball, the average error was 280 mm and the maximum was ~600 mm. In this study, we investigated a method for more precisely estimating the drop position of a ball that overcomes the effects of lighting. We photographed table tennis balls using a single high-speed camera and estimated their drop position by combining a background difference method and a particle filter. As a result, the errors between the actual and estimated drop positions of balls was an average of 40 mm to 60 mm, and the estimation precision was improved because the incorrect estimates were significantly reduced.
© Copyright 2020 Proceedings Book of The 16th ITTF Sports Science Congress Budapest (HUN), April 19th - 20th, 2019. Published by International Table Tennis Federation. All rights reserved.