Analysis of Playing Positions in Tennis Match Videos to Assess Competition Using a Centroid Clustering Heatmap Prediction Technique

Abstract

This research aimed to use clustered heatmap positioning analytical techniques in tennis in order to be able to analyze the positions of tennis players. A heatmap represents the cumulative frequency of tennis players’ movements in each zone of the tennis court. The performance testing of centroid clustering heatmap position analysis was achieved on selected men’s doubles tennis matches during the SINGHA CLASSIC 2019 competition. The research was done by collecting the cumulative frequency data and replacing it with intensity of color space. The process started with, firstly, cutting videos for each match based on the area of the court that could be seen clearly by the cameras in the field. Secondly, the video was converted into binary images. Thirdly, noise reduction was performed using morphological techniques. Fourthly, the centroid position was identified using a connected component and blob analysis. Fifthly, clustering data with k-mine was used to predict new tracks by Kalman filter. Finally, the percentage of player position in the three zones of the tennis court was calculated with the percent yield formula. The experimental results clearly showed the cumulative frequency of the players’ movement with the intensity of color space, allowing coaches and players to easily understand and use the data in planning for the next practice or competition.