LSTM-Based Prediction of Ball Trajectory for Automatic Foul Judgements in Kin-Ball Sport

Video analysis of sports has become an important means of analyzing the judgment of foul play and performance in matches. Traditionally, most researches have focuses on major sports, but few researches on minor sports have progressed. Therefore, this paper focuses on a minor sport called Kin-Ball and proposes a system for judging fouls in Kin-Ball. One of the fouls in Kin-Ball is called “downward hits,” which is a foul if the trajectory of the ball is not upward at least once from when the ball is hit and to when the ball hits the ground during an attack to the other team. This foul is a feature that does not exist in other sports, and it is very difficult to judge visually by human referees as well, their decision may be ambiguous. Therefore, this paper proposes a novel system for CNN and LSTM-based predicting the trajectory of the ball in Kin-Ball play video taken with a single camera, and automatically judging “downward hits” fouls based on its predicted trajectory.