Game analysis and prediction of ball positions in a football match from video footages

Abstract

In this paper, we have proposed two algorithms respectively for automatic analysis and predictions for most optimum passes in the football match using the detection and tracking of players as well as of ball from the video footage. Using our first algorithm initially checked for the valid frame using grass ratio and based on this ratio, the playing field is marked for each frame. Then algorithm performs the detection and tracking using the connectivity table automatically find possession of a ball for the both team during the match. Finally, we also perform the replay checks, which are usually shown after every good move or goal scored, using the background check to find the possible deviation in player position in last few seconds of clipping. Another algorithm is developed for the prediction of optimum pass by an individual during game. Our algorithm first finds the player who possessed the ball and which team he belongs. In second step, search for the nearest teammate using the heuristic distance. In third step, algorithm searches whether there is any player from the opposition is obstructing or closes enough to obstructing a line of possible pass. In the last step, algorithm checks whether the player selected for delivering a pass is surrounded by opposition team in a close proximity. This process will repeat till he fined a free teammate for ball passing.